SI_2_multiplicative_vs_additive.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as mpatches\n",
    "%matplotlib inline\n",
    "plt.rc(\"axes.spines\", top=False, right=False)\n",
    "\n",
    "from plotting_functions import plot_correlation, add_regression_line, cm2inch, show_img\n",
    "from analysis_functions import compute_gaze_influence_score, compute_mean_rt, compute_p_choose_best, add_best_chosen, aggregate_subject_level\n",
    "from analysis_functions import std, se, q1, q3, iqr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.random.seed(84) # from random.org"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2.2 Relative model fit (individual level model comparison)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load WAICs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>subject</th>\n",
       "      <th>dWAIC</th>\n",
       "      <th>dSE</th>\n",
       "      <th>dataset</th>\n",
       "      <th>warning</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>-0.90</td>\n",
       "      <td>0.99</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>-1.19</td>\n",
       "      <td>3.06</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>1.52</td>\n",
       "      <td>1.52</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>-0.65</td>\n",
       "      <td>1.58</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>-3.16</td>\n",
       "      <td>1.79</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   subject  dWAIC   dSE       dataset  warning\n",
       "0        0  -0.90  0.99  krajbich2010      0.0\n",
       "1        1  -1.19  3.06  krajbich2010      0.0\n",
       "2        2   1.52  1.52  krajbich2010      0.0\n",
       "3        3  -0.65  1.58  krajbich2010      0.0\n",
       "4        4  -3.16  1.79  krajbich2010      0.0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "waic_summaries = [file for file in os.listdir(os.path.join('results', 'model_comparison', 'additive_vs_multiplicative'))\n",
    "                  if file.endswith('waic.csv')]\n",
    "\n",
    "n_krajbich2010 = 39\n",
    "n_krajbich2011 = 30\n",
    "n_folke2016 = 24\n",
    "n_tavares2017 = 25\n",
    "n = n_krajbich2010 + n_krajbich2011 + n_folke2016 + n_tavares2017\n",
    "\n",
    "waic = pd.DataFrame(dict(subject=np.arange(n)))\n",
    "waic['dWAIC'] = np.nan\n",
    "waic['dSE'] = np.nan\n",
    "waic['dataset'] = np.nan\n",
    "waic['warning'] = np.nan\n",
    "\n",
    "for file in waic_summaries:\n",
    "    _, _, _, subject, _ = file.split('_')\n",
    "    subject = int(subject)\n",
    "\n",
    "    waic_summary = pd.read_csv(os.path.join('results', 'model_comparison',\n",
    "                                            'additive_vs_multiplicative', file), index_col='Unnamed: 0')\n",
    "    waic.loc[waic['subject'] == subject, 'dWAIC'] = waic_summary.loc['GLAM', 'WAIC'] - waic_summary.loc['additive', 'WAIC']\n",
    "    waic.loc[waic['subject'] == subject, 'dSE'] = waic_summary['dSE'].values[1]\n",
    "    waic.loc[waic['subject'] == subject, 'warning'] = waic_summary['var_warn'].sum()\n",
    "\n",
    "waic.sort_values('subject', inplace=True)\n",
    "waic['dataset'] = np.array(n_krajbich2010 * ['krajbich2010'] + n_krajbich2011 * ['krajbich2011'] + n_folke2016 * ['folke2016'] + n_tavares2017 * ['tavares2017'])\n",
    "\n",
    "waic.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plot WAIC differences between multiplicative and additive GLAM variants"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Panel a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def figure_si_2a(waic, colors=None, fontsize=7, figsize=cm2inch((18, 6)), plot_type='bar', log_scale=False, ylim=(-50, 100), plot_legend=True, ax=None):\n",
    "\n",
    "    # Log transform: add minimum dWAIC + 1, then log-transform, new indifference then at log(minimum dWAIC + 1)\n",
    "    waic = waic.copy()\n",
    "    if log_scale:\n",
    "        min_dWAIC = np.min(waic['dWAIC'])\n",
    "        indifference_level = np.log(np.abs(min_dWAIC) + 1)\n",
    "        waic['plotted_dWAIC'] = np.log(waic['dWAIC'] + np.abs(min_dWAIC) + 1)\n",
    "    else:\n",
    "        indifference_level = 0\n",
    "        waic['plotted_dWAIC'] = waic['dWAIC']\n",
    "    \n",
    "    if colors is None:\n",
    "        colors = np.array(['C{}'.format(i) for i in range(10)])\n",
    "\n",
    "    datasets = ['krajbich2010', 'krajbich2011', 'folke2016', 'tavares2017']\n",
    "    dataset_labels = ['Krajbich 2010', 'Krajbich 2011', 'Folke 2016', 'Tavares 2017']\n",
    "    \n",
    "    if ax is None:\n",
    "        fig, ax = plt.subplots(figsize=figsize)\n",
    "\n",
    "    sorting_idx = np.argsort(waic['plotted_dWAIC'])\n",
    "    color_idx = pd.Categorical(waic['dataset'], categories=datasets).codes\n",
    "    patches = []\n",
    "    \n",
    "    patches = [mpatches.Rectangle((0, 0), 0.5, 1, fc=\"C{}\".format(i))\n",
    "               for i in range(len(datasets))]\n",
    "\n",
    "    for i, dataset in enumerate(datasets):\n",
    "\n",
    "        if plot_type == 'bar':\n",
    "            ax.bar(np.arange(len(waic))[waic['dataset'][sorting_idx] == dataset],\n",
    "                   waic['plotted_dWAIC'][sorting_idx][waic['dataset'][sorting_idx] == dataset],\n",
    "                   color=colors[i])\n",
    "\n",
    "        else:\n",
    "            ax.scatter(np.arange(len(waic))[waic['dataset'][sorting_idx] == dataset],\n",
    "                       waic['plotted_dWAIC'][sorting_idx][waic['dataset'][sorting_idx] == dataset],\n",
    "                       color=colors[i], marker='+')\n",
    "\n",
    "    \n",
    "    # Legend\n",
    "    if plot_legend:\n",
    "        ax.legend(patches,\n",
    "                  ['Krajbich 2010', 'Krajbich 2011', 'Folke 2016', 'Tavares 2017'],\n",
    "                  ncol=2,\n",
    "                  loc='lower right',\n",
    "                  handlelength=1, handleheight=1,\n",
    "                  fontsize=fontsize, columnspacing=1,\n",
    "                  frameon=False)\n",
    "    \n",
    "    \n",
    "    ax.set_xlabel('Participant', fontsize=fontsize)\n",
    "    if log_scale:\n",
    "        ax.set_ylabel('Log(dWAIC))', fontsize=fontsize)\n",
    "    else:\n",
    "        ax.set_ylabel('WAIC(multiplicative)-\\nWAIC(additive)', fontsize=fontsize)\n",
    "    ax.set_xticks([])\n",
    "    ax.tick_params(axis='both', which='major', labelsize=fontsize)\n",
    "    \n",
    "    ax.axhline(indifference_level, color='black', lw=1, zorder=9)\n",
    "    \n",
    "    ax.set_ylim(ylim)\n",
    "    \n",
    "    \n",
    "    ax.text(-0.075, 1., 'a', transform=ax.transAxes,\n",
    "                fontsize=7, fontweight='bold', va='top')\n",
    "    ax.spines['top'].set_visible(False)\n",
    "    ax.spines['right'].set_visible(False)\n",
    "        \n",
    "    # Font sizes\n",
    "    ax.tick_params(axis='both', labelsize=8)\n",
    "    \n",
    "    return ax"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcIAAACcCAYAAADs66t2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\na0qEIiKS1pQIRUQkrSkRiohIWvt/2tWHIfM+iGYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x104f30a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "figure_si_2a(waic, ylim=(-5, 5));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    118.000000\n",
       "mean       2.100424\n",
       "std       11.626632\n",
       "min      -35.910000\n",
       "25%       -1.605000\n",
       "50%        0.720000\n",
       "75%        5.157500\n",
       "max       56.600000\n",
       "Name: dWAIC, dtype: float64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "waic['dWAIC'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "68"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(waic['dWAIC'] > 0).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.576271186440678"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(waic['dWAIC'] > 0).sum() / len(waic)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Paneld bcd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "all_data = pd.read_csv('data/data_aggregate.csv')\n",
    "even = all_data[(all_data['trial'] % 2) == 0].reset_index(drop=True)\n",
    "odd = all_data[(all_data['trial'] % 2) == 1].reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/armin/anaconda3/lib/python3.5/site-packages/ipykernel/__main__.py:15: FutureWarning: Sorting because non-concatenation axis is not aligned. A future version\n",
      "of pandas will change to not sort by default.\n",
      "\n",
      "To accept the future behavior, pass 'sort=False'.\n",
      "\n",
      "To retain the current behavior and silence the warning, pass 'sort=True'.\n",
      "\n",
      "/Users/armin/anaconda3/lib/python3.5/site-packages/ipykernel/__main__.py:16: FutureWarning: Sorting because non-concatenation axis is not aligned. A future version\n",
      "of pandas will change to not sort by default.\n",
      "\n",
      "To accept the future behavior, pass 'sort=False'.\n",
      "\n",
      "To retain the current behavior and silence the warning, pass 'sort=True'.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# read model predictions for multiplicative and additive variants and concatenate\n",
    "prediction_additive_list = []\n",
    "prediction_multiplicative_list = []\n",
    "\n",
    "# join prediction dataframes\n",
    "for subject in odd['subject'].unique():\n",
    "    subject_pred_additive = pd.read_csv(os.path.join('results', 'predictions', 'out_of_sample', 'additive', 'prediction_{}_additive_oos.csv'.format(subject)), index_col=0)\n",
    "    subject_pred_additive['subject'] = subject    \n",
    "    prediction_additive_list.append(subject_pred_additive)\n",
    "\n",
    "    subject_pred_multiplicative = pd.read_csv(os.path.join('results', 'predictions', 'out_of_sample', 'multiplicative', 'prediction_{}_multiplicative_oos.csv'.format(subject)), index_col=0)\n",
    "    subject_pred_multiplicative['subject'] = subject\n",
    "    prediction_multiplicative_list.append(subject_pred_multiplicative)\n",
    "\n",
    "prediction_additive = pd.concat(prediction_additive_list)\n",
    "prediction_multiplicative = pd.concat(prediction_multiplicative_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# discard trials with negative predicted RT\n",
    "prediction_additive = prediction_additive[prediction_additive['rt'] > 0].copy()\n",
    "prediction_multiplicative = prediction_multiplicative[prediction_multiplicative['rt'] > 0].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# compute / load individual behavioural measures for odd data and model predictions\n",
    "odd_summary = pd.DataFrame()\n",
    "data_summary = pd.DataFrame()\n",
    "pred_additive_summary = pd.DataFrame()\n",
    "\n",
    "for di, dataset in enumerate(odd['dataset'].unique()):\n",
    "    odd_di = odd[odd['dataset'] == dataset].copy()\n",
    "    n_items = odd_di['n_items'].unique()[0]\n",
    "    \n",
    "    odd_summary_di = aggregate_subject_level(odd_di, n_items=n_items)\n",
    "    \n",
    "    prediction_additive_di = prediction_additive[prediction_additive['dataset'] == dataset].copy()\n",
    "    \n",
    "    pred_additive_summary_di = aggregate_subject_level(prediction_additive_di, n_items=n_items)\n",
    "    \n",
    "    \n",
    "    odd_summary = pd.concat([odd_summary, odd_summary_di])\n",
    "    pred_additive_summary = pd.concat([pred_additive_summary, pred_additive_summary_di])\n",
    "\n",
    "pred_additive_summary.to_csv(os.path.join('results', 'predictions', 'out_of_sample', 'descriptives',\n",
    "                                          'subject_summary_additive_oos.csv'))\n",
    "\n",
    "pred_multiplicative_summary = pd.read_csv(os.path.join('results',\n",
    "                                                       'predictions',\n",
    "                                                       'out_of_sample',\n",
    "                                                       'descriptives',\n",
    "                                                       'subject_summary_multiplicative_oos.csv'),\n",
    "                                         header=[0, 1], index_col=0)\n",
    "pred_additive_summary = pd.read_csv(os.path.join('results',\n",
    "                                                 'predictions',\n",
    "                                                 'out_of_sample',\n",
    "                                                 'descriptives',\n",
    "                                                 'subject_summary_additive_oos.csv'),\n",
    "                                    header=[0, 1], index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def figure_si_2bcd(multiplicative, additive,\n",
    "                 figsize=cm2inch(18, 6), colors=None, fontsize=7, alpha=0.5, axs=None, plot_legend=False):\n",
    "\n",
    "    datasets = ['krajbich2010', 'krajbich2011', 'folke2016', 'tavares2017']\n",
    "\n",
    "    if colors is None:\n",
    "        colors = np.array(['C{}'.format(i)\n",
    "                           for i in range(multiplicative['dataset']['Unnamed: 11_level_1'].unique().size)])\n",
    "    color_idx = pd.Categorical(multiplicative['dataset']['Unnamed: 11_level_1'], categories=datasets, ordered=True).codes.astype(int)\n",
    "    \n",
    "    if axs is None:\n",
    "        fig, axs = plt.subplots(1, 3, figsize=figsize, sharex='col', sharey='col')\n",
    "\n",
    "    # b) Mean RT\n",
    "    axs[0].scatter(multiplicative['rt']['mean'],\n",
    "                   additive['rt']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   linewidth=0,\n",
    "                   marker='o',\n",
    "                   alpha=alpha)\n",
    "    axs[0].scatter(multiplicative['rt']['mean'],\n",
    "                   additive['rt']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   facecolor='none',\n",
    "                   linewidth=0.5,\n",
    "                   marker='o',\n",
    "                   alpha=1.0)\n",
    "    axs[0].set_xticks(np.arange(0, 10000, 2000))\n",
    "    axs[0].set_yticks(np.arange(0, 10000, 2000))\n",
    "    axs[0].set_xlim(0, 9000)\n",
    "    axs[0].set_ylim(0, 9000)\n",
    "\n",
    "\n",
    "    # c) P(choose best)\n",
    "    axs[1].scatter(multiplicative['best_chosen']['mean'],\n",
    "                   additive['best_chosen']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   linewidth=0,\n",
    "                   marker='o',\n",
    "                   alpha=alpha)\n",
    "    axs[1].scatter(multiplicative['best_chosen']['mean'],\n",
    "                   additive['best_chosen']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   facecolor='none',\n",
    "                   linewidth=0.5,\n",
    "                   marker='o',\n",
    "                   alpha=1.0)\n",
    "\n",
    "    axs[1].set_xticks(np.arange(0.4, 1.1, 0.2))\n",
    "    axs[1].set_yticks(np.arange(0.4, 1.1, 0.2))\n",
    "    axs[1].set_xlim(0.2, 1.0)\n",
    "    axs[1].set_ylim(0.2, 1.0)\n",
    "\n",
    "\n",
    "\n",
    "    # d) Gaze Influence\n",
    "    axs[2].scatter(multiplicative['gaze_influence']['Unnamed: 10_level_1'],\n",
    "                   additive['gaze_influence']['Unnamed: 10_level_1'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   linewidth=0,\n",
    "                   alpha=alpha)\n",
    "    axs[2].scatter(multiplicative['gaze_influence']['Unnamed: 10_level_1'],\n",
    "                   additive['gaze_influence']['Unnamed: 10_level_1'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   facecolor='none',\n",
    "                   linewidth=0.5,\n",
    "                   alpha=1.0)\n",
    "    \n",
    "    axs[2].set_xticks(np.arange(-0.2, 1.1, 0.2))\n",
    "    axs[2].set_yticks(np.arange(-0.2, 1.1, 0.2))\n",
    "    axs[2].set_xlim(-0.3, 0.85)\n",
    "    axs[2].set_ylim(-0.3, 0.85)\n",
    "\n",
    "    axs[0].set_ylabel('Additive variant predicted\\nMean RT (ms)', fontsize=fontsize)\n",
    "    axs[0].set_xlabel('Multiplicative variant predicted\\nMean RT (ms)', fontsize=fontsize)\n",
    "\n",
    "    axs[1].set_ylabel('Additive variant predicted\\nP(choose best)', fontsize=fontsize)\n",
    "    axs[1].set_xlabel('Multiplicative variant predicted\\nP(choose best)', fontsize=fontsize)\n",
    "\n",
    "    axs[2].set_ylabel('Additive variant predicted\\nGaze influence', fontsize=fontsize)\n",
    "    axs[2].set_xlabel('Multiplicative variant predicted\\nGaze influence', fontsize=fontsize)\n",
    "    \n",
    "    patches = [mpatches.Rectangle((0, 0), 0.5, 1, fc=\"C{}\".format(i))\n",
    "               for i in range(len(datasets))]\n",
    "    if plot_legend:\n",
    "        axs[0].legend(patches,\n",
    "                         ['Krajbich 2010', 'Krajbich 2011', 'Folke 2016', 'Tavares 2017'],\n",
    "                         loc='upper left',\n",
    "                         handlelength=1, handleheight=1,\n",
    "                         fontsize=fontsize,\n",
    "                         frameon=False)\n",
    "\n",
    "    # Labels\n",
    "    for label, ax in zip(list('bcd'), axs.ravel()):\n",
    "        ax.text(-0.33, 1.1, label, transform=ax.transAxes,\n",
    "                fontsize=fontsize, fontweight='bold', va='top')\n",
    "        ax.spines['top'].set_visible(False)\n",
    "        ax.spines['right'].set_visible(False)\n",
    "        \n",
    "        # Font sizes\n",
    "        ax.tick_params(axis='both', labelsize=fontsize)\n",
    "\n",
    "        # Plot diagonal\n",
    "        ax.plot(ax.get_xlim(), ax.get_xlim(), color='black', linewidth=1, alpha=1.0, zorder=-1)\n",
    "            \n",
    "    return axs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Panels efg (additive model parameters vs. behavioural measures) – in sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>best_chosen</th>\n",
       "      <th colspan=\"8\" halign=\"left\">rt</th>\n",
       "      <th>gaze_influence</th>\n",
       "      <th>dataset</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>max</th>\n",
       "      <th>se</th>\n",
       "      <th>q1</th>\n",
       "      <th>q3</th>\n",
       "      <th>iqr</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>subject</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.822222</td>\n",
       "      <td>1704.177778</td>\n",
       "      <td>881.852300</td>\n",
       "      <td>682.0</td>\n",
       "      <td>4142.0</td>\n",
       "      <td>132.944236</td>\n",
       "      <td>1125.00</td>\n",
       "      <td>1970.0</td>\n",
       "      <td>845.00</td>\n",
       "      <td>-0.020529</td>\n",
       "      <td>krajbich2010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.673469</td>\n",
       "      <td>1728.489796</td>\n",
       "      <td>857.141479</td>\n",
       "      <td>743.0</td>\n",
       "      <td>4602.0</td>\n",
       "      <td>123.717716</td>\n",
       "      <td>1166.00</td>\n",
       "      <td>1937.0</td>\n",
       "      <td>771.00</td>\n",
       "      <td>0.247233</td>\n",
       "      <td>krajbich2010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.840000</td>\n",
       "      <td>4078.660000</td>\n",
       "      <td>2623.044800</td>\n",
       "      <td>943.0</td>\n",
       "      <td>12079.0</td>\n",
       "      <td>374.720686</td>\n",
       "      <td>2104.25</td>\n",
       "      <td>5329.0</td>\n",
       "      <td>3224.75</td>\n",
       "      <td>0.028877</td>\n",
       "      <td>krajbich2010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.704545</td>\n",
       "      <td>1893.545455</td>\n",
       "      <td>633.499496</td>\n",
       "      <td>1045.0</td>\n",
       "      <td>4742.0</td>\n",
       "      <td>96.607767</td>\n",
       "      <td>1436.50</td>\n",
       "      <td>2175.5</td>\n",
       "      <td>739.00</td>\n",
       "      <td>-0.022031</td>\n",
       "      <td>krajbich2010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.840000</td>\n",
       "      <td>1846.640000</td>\n",
       "      <td>1093.485579</td>\n",
       "      <td>692.0</td>\n",
       "      <td>5891.0</td>\n",
       "      <td>156.212226</td>\n",
       "      <td>1221.50</td>\n",
       "      <td>2132.5</td>\n",
       "      <td>911.00</td>\n",
       "      <td>0.103622</td>\n",
       "      <td>krajbich2010</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        best_chosen           rt                                            \\\n",
       "               mean         mean          std     min      max          se   \n",
       "subject                                                                      \n",
       "0          0.822222  1704.177778   881.852300   682.0   4142.0  132.944236   \n",
       "1          0.673469  1728.489796   857.141479   743.0   4602.0  123.717716   \n",
       "2          0.840000  4078.660000  2623.044800   943.0  12079.0  374.720686   \n",
       "3          0.704545  1893.545455   633.499496  1045.0   4742.0   96.607767   \n",
       "4          0.840000  1846.640000  1093.485579   692.0   5891.0  156.212226   \n",
       "\n",
       "                                  gaze_influence       dataset  \n",
       "              q1      q3      iqr                               \n",
       "subject                                                         \n",
       "0        1125.00  1970.0   845.00      -0.020529  krajbich2010  \n",
       "1        1166.00  1937.0   771.00       0.247233  krajbich2010  \n",
       "2        2104.25  5329.0  3224.75       0.028877  krajbich2010  \n",
       "3        1436.50  2175.5   739.00      -0.022031  krajbich2010  \n",
       "4        1221.50  2132.5   911.00       0.103622  krajbich2010  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "odd_summary.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>gamma</th>\n",
       "      <th>s</th>\n",
       "      <th>subject</th>\n",
       "      <th>tau</th>\n",
       "      <th>v</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0.0</th>\n",
       "      <td>1.78</td>\n",
       "      <td>0.008020</td>\n",
       "      <td>0</td>\n",
       "      <td>0.36</td>\n",
       "      <td>0.000096</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <td>7.87</td>\n",
       "      <td>0.010117</td>\n",
       "      <td>1</td>\n",
       "      <td>0.10</td>\n",
       "      <td>0.000100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <td>0.49</td>\n",
       "      <td>0.008944</td>\n",
       "      <td>2</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.000023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <td>1.11</td>\n",
       "      <td>0.004812</td>\n",
       "      <td>3</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.000089</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4.0</th>\n",
       "      <td>6.00</td>\n",
       "      <td>0.009909</td>\n",
       "      <td>4</td>\n",
       "      <td>0.20</td>\n",
       "      <td>0.000092</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     gamma         s  subject   tau         v\n",
       "0.0   1.78  0.008020        0  0.36  0.000096\n",
       "1.0   7.87  0.010117        1  0.10  0.000100\n",
       "2.0   0.49  0.008944        2  1.00  0.000023\n",
       "3.0   1.11  0.004812        3  0.01  0.000089\n",
       "4.0   6.00  0.009909        4  0.20  0.000092"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "estimates = pd.DataFrame()\n",
    "parameter_names = ['v', 'gamma', 's', 'tau']\n",
    "for subject in odd_summary.index:\n",
    "    subject_estimates_df = pd.read_csv(os.path.join('results', 'estimates', 'out_of_sample', 'additive',\n",
    "                                                    'estimates_{}_additive_oos.csv'.format(subject)),\n",
    "                                       index_col=0)\n",
    "    subject_estimates_dict = {parameter: subject_estimates_df.loc[parameter + '__0_0', 'MAP']\n",
    "                              for parameter in parameter_names}\n",
    "    subject_estimates_dict['subject'] = subject\n",
    "    estimates = pd.concat([estimates,\n",
    "                           pd.DataFrame(subject_estimates_dict,\n",
    "                                        index=np.ones(1)*subject)])\n",
    "\n",
    "estimates.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def figure_si_2efg(estimates, summary,\n",
    "                   figsize=cm2inch(18, 6), colors=None, fontsize=7, alpha=0.5, axs=None, plot_legend=False):\n",
    "\n",
    "    datasets = ['krajbich2010', 'krajbich2011', 'folke2016', 'tavares2017']\n",
    "\n",
    "    if colors is None:\n",
    "        colors = np.array(['C{}'.format(i)\n",
    "                           for i in range(summary['dataset'].unique().size)])\n",
    "    color_idx = pd.Categorical(summary['dataset'], categories=datasets, ordered=True).codes.astype(int)\n",
    "    \n",
    "    if axs is None:\n",
    "        fig, axs = plt.subplots(1, 3, figsize=figsize, sharex='col', sharey='col')\n",
    "\n",
    "    # e) log Mean RT ~ log v\n",
    "    axs[0].scatter(np.log(estimates['v']),\n",
    "                   np.log(summary['rt']['mean']),\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   linewidth=0,\n",
    "                   alpha=alpha)\n",
    "    axs[0].scatter(np.log(estimates['v']),\n",
    "                   np.log(summary['rt']['mean']),\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   facecolor='none',\n",
    "                   linewidth=0.5,\n",
    "                   alpha=1.0)\n",
    "    axs[0].set_ylim(6, 9.125)\n",
    "    \n",
    "    # b) P(choose best) ~ gamma\n",
    "    axs[1].scatter(estimates['gamma'],\n",
    "                   summary['best_chosen']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   linewidth=0,\n",
    "                   alpha=alpha)\n",
    "    axs[1].scatter(estimates['gamma'],\n",
    "                   summary['best_chosen']['mean'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   facecolor='none',\n",
    "                   linewidth=0.5,\n",
    "                   alpha=1.0)\n",
    "    axs[1].set_ylim(0.3, 1.0)\n",
    "\n",
    "    # c) Gaze Influence ~ gamma\n",
    "    axs[2].scatter(estimates['gamma'],\n",
    "                   summary['gaze_influence'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   linewidth=0,\n",
    "                   alpha=alpha)\n",
    "    axs[2].scatter(estimates['gamma'],\n",
    "                   summary['gaze_influence'],\n",
    "                   color=colors[color_idx],\n",
    "                   marker='o',\n",
    "                   linewidth=0.5,\n",
    "                   facecolor='none',\n",
    "                   alpha=1.0)\n",
    "\n",
    "    axs[0].set_ylabel('Log mean RT (ms)', fontsize=fontsize)\n",
    "    axs[1].set_ylabel('P(choose best)', fontsize=fontsize)\n",
    "    axs[2].set_ylabel('Gaze influence\\non P(choice | value)', fontsize=fontsize)\n",
    "     \n",
    "    axs[0].set_xlabel(r'Log $\\hat{v}$', fontsize=fontsize)\n",
    "    axs[1].set_xlabel(r'$\\hat{\\gamma}_{additive}$', fontsize=fontsize)\n",
    "    axs[2].set_xlabel(r'$\\hat{\\gamma}_{additive}$', fontsize=fontsize)\n",
    "    \n",
    "    patches = [mpatches.Rectangle((0, 0), 0.5, 1, fc=\"C{}\".format(i))\n",
    "               for i in range(len(datasets))]\n",
    "    if plot_legend:\n",
    "        axs[0].legend(patches,\n",
    "                         ['Krajbich 2010', 'Krajbich 2011', 'Folke 2016', 'Tavares 2017'],\n",
    "                         loc='upper left',\n",
    "                         handlelength=1, handleheight=1,\n",
    "                         fontsize=fontsize,\n",
    "                         frameon=False)\n",
    "\n",
    "    # Labels\n",
    "    for label, ax in zip(list('efg'), axs.ravel()):\n",
    "        ax.text(-0.3, 1.1, label, transform=ax.transAxes,\n",
    "                fontsize=fontsize, fontweight='bold', va='top')\n",
    "        ax.spines['top'].set_visible(False)\n",
    "        ax.spines['right'].set_visible(False)\n",
    "        \n",
    "        # Font sizes\n",
    "        ax.tick_params(axis='both', labelsize=fontsize)\n",
    "            \n",
    "    return axs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Mixed effects regressions with random slope and intercept for data sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from bambi import Model\n",
    "from pymc3 import plot_posterior"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Format data for mixed model\n",
    "df = pd.concat([estimates, odd_summary], axis=1)\n",
    "df['log_rt'] = np.log(df[('rt', 'mean')])\n",
    "df['log_v'] = np.log(df['v'])\n",
    "df['best_chosen'] = 100 * df[('best_chosen', 'mean')]\n",
    "df['gaze_influence'] = 100 * df[('gaze_influence', '')]\n",
    "df['dataset'] = df[('dataset', '')]\n",
    "\n",
    "df = df[['subject', 'dataset', 'log_rt', 'best_chosen', 'gaze_influence', 'log_v', 'gamma', 's', 'tau']].copy().reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>subject</th>\n",
       "      <th>dataset</th>\n",
       "      <th>log_rt</th>\n",
       "      <th>best_chosen</th>\n",
       "      <th>gaze_influence</th>\n",
       "      <th>log_v</th>\n",
       "      <th>gamma</th>\n",
       "      <th>s</th>\n",
       "      <th>tau</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>7.440838</td>\n",
       "      <td>82.222222</td>\n",
       "      <td>-2.052946</td>\n",
       "      <td>-9.251162</td>\n",
       "      <td>1.78</td>\n",
       "      <td>0.008020</td>\n",
       "      <td>0.36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>7.455003</td>\n",
       "      <td>67.346939</td>\n",
       "      <td>24.723267</td>\n",
       "      <td>-9.210340</td>\n",
       "      <td>7.87</td>\n",
       "      <td>0.010117</td>\n",
       "      <td>0.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>8.313524</td>\n",
       "      <td>84.000000</td>\n",
       "      <td>2.887728</td>\n",
       "      <td>-10.680016</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.008944</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>7.546206</td>\n",
       "      <td>70.454545</td>\n",
       "      <td>-2.203105</td>\n",
       "      <td>-9.326874</td>\n",
       "      <td>1.11</td>\n",
       "      <td>0.004812</td>\n",
       "      <td>0.01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>7.521123</td>\n",
       "      <td>84.000000</td>\n",
       "      <td>10.362165</td>\n",
       "      <td>-9.293722</td>\n",
       "      <td>6.00</td>\n",
       "      <td>0.009909</td>\n",
       "      <td>0.20</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   subject       dataset    log_rt  best_chosen  gaze_influence      log_v  \\\n",
       "0        0  krajbich2010  7.440838    82.222222       -2.052946  -9.251162   \n",
       "1        1  krajbich2010  7.455003    67.346939       24.723267  -9.210340   \n",
       "2        2  krajbich2010  8.313524    84.000000        2.887728 -10.680016   \n",
       "3        3  krajbich2010  7.546206    70.454545       -2.203105  -9.326874   \n",
       "4        4  krajbich2010  7.521123    84.000000       10.362165  -9.293722   \n",
       "\n",
       "   gamma         s   tau  \n",
       "0   1.78  0.008020  0.36  \n",
       "1   7.87  0.010117  0.10  \n",
       "2   0.49  0.008944  1.00  \n",
       "3   1.11  0.004812  0.01  \n",
       "4   6.00  0.009909  0.20  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "n_samples = 20000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading previous results from \"results/mixed_effects_models/si2_parameter_behavior_e_v_rt.csv\"...\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>-0.017646</td>\n",
       "      <td>0.295195</td>\n",
       "      <td>-0.634416</td>\n",
       "      <td>0.423502</td>\n",
       "      <td>3525</td>\n",
       "      <td>1.004311</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>0.033198</td>\n",
       "      <td>0.243402</td>\n",
       "      <td>-0.416619</td>\n",
       "      <td>0.590320</td>\n",
       "      <td>6036</td>\n",
       "      <td>1.000039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>-0.125765</td>\n",
       "      <td>0.286000</td>\n",
       "      <td>-0.728672</td>\n",
       "      <td>0.257428</td>\n",
       "      <td>2929</td>\n",
       "      <td>1.004666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>-0.071335</td>\n",
       "      <td>0.278042</td>\n",
       "      <td>-0.640520</td>\n",
       "      <td>0.357967</td>\n",
       "      <td>3211</td>\n",
       "      <td>1.005446</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>0.231486</td>\n",
       "      <td>0.306003</td>\n",
       "      <td>0.000026</td>\n",
       "      <td>0.828555</td>\n",
       "      <td>2254</td>\n",
       "      <td>1.008188</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>-0.033978</td>\n",
       "      <td>0.354079</td>\n",
       "      <td>-0.689633</td>\n",
       "      <td>0.629774</td>\n",
       "      <td>3947</td>\n",
       "      <td>1.000531</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_rt_sd</th>\n",
       "      <td>0.124105</td>\n",
       "      <td>0.008430</td>\n",
       "      <td>0.108121</td>\n",
       "      <td>0.140994</td>\n",
       "      <td>31014</td>\n",
       "      <td>1.000250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v</th>\n",
       "      <td>-0.798045</td>\n",
       "      <td>0.038026</td>\n",
       "      <td>-0.868558</td>\n",
       "      <td>-0.728922</td>\n",
       "      <td>3577</td>\n",
       "      <td>1.001977</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v|dataset[folke2016]</th>\n",
       "      <td>-0.012266</td>\n",
       "      <td>0.032372</td>\n",
       "      <td>-0.077977</td>\n",
       "      <td>0.032816</td>\n",
       "      <td>2263</td>\n",
       "      <td>1.007202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v|dataset[krajbich2010]</th>\n",
       "      <td>0.003156</td>\n",
       "      <td>0.026932</td>\n",
       "      <td>-0.042481</td>\n",
       "      <td>0.064316</td>\n",
       "      <td>4049</td>\n",
       "      <td>1.000501</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v|dataset[krajbich2011]</th>\n",
       "      <td>-0.005552</td>\n",
       "      <td>0.032312</td>\n",
       "      <td>-0.072345</td>\n",
       "      <td>0.037952</td>\n",
       "      <td>1968</td>\n",
       "      <td>1.007821</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v|dataset[tavares2017]</th>\n",
       "      <td>-0.007439</td>\n",
       "      <td>0.032203</td>\n",
       "      <td>-0.074850</td>\n",
       "      <td>0.041168</td>\n",
       "      <td>2100</td>\n",
       "      <td>1.008435</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>log_v|dataset_sd</th>\n",
       "      <td>0.025354</td>\n",
       "      <td>0.038486</td>\n",
       "      <td>0.000001</td>\n",
       "      <td>0.093527</td>\n",
       "      <td>1467</td>\n",
       "      <td>1.012255</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 mean        sd  hpd0.95_lower  hpd0.95_upper  \\\n",
       "1|dataset[folke2016]        -0.017646  0.295195      -0.634416       0.423502   \n",
       "1|dataset[krajbich2010]      0.033198  0.243402      -0.416619       0.590320   \n",
       "1|dataset[krajbich2011]     -0.125765  0.286000      -0.728672       0.257428   \n",
       "1|dataset[tavares2017]      -0.071335  0.278042      -0.640520       0.357967   \n",
       "1|dataset_sd                 0.231486  0.306003       0.000026       0.828555   \n",
       "Intercept                   -0.033978  0.354079      -0.689633       0.629774   \n",
       "log_rt_sd                    0.124105  0.008430       0.108121       0.140994   \n",
       "log_v                       -0.798045  0.038026      -0.868558      -0.728922   \n",
       "log_v|dataset[folke2016]    -0.012266  0.032372      -0.077977       0.032816   \n",
       "log_v|dataset[krajbich2010]  0.003156  0.026932      -0.042481       0.064316   \n",
       "log_v|dataset[krajbich2011] -0.005552  0.032312      -0.072345       0.037952   \n",
       "log_v|dataset[tavares2017]  -0.007439  0.032203      -0.074850       0.041168   \n",
       "log_v|dataset_sd             0.025354  0.038486       0.000001       0.093527   \n",
       "\n",
       "                             effective_n  gelman_rubin  \n",
       "1|dataset[folke2016]                3525      1.004311  \n",
       "1|dataset[krajbich2010]             6036      1.000039  \n",
       "1|dataset[krajbich2011]             2929      1.004666  \n",
       "1|dataset[tavares2017]              3211      1.005446  \n",
       "1|dataset_sd                        2254      1.008188  \n",
       "Intercept                           3947      1.000531  \n",
       "log_rt_sd                          31014      1.000250  \n",
       "log_v                               3577      1.001977  \n",
       "log_v|dataset[folke2016]            2263      1.007202  \n",
       "log_v|dataset[krajbich2010]         4049      1.000501  \n",
       "log_v|dataset[krajbich2011]         1968      1.007821  \n",
       "log_v|dataset[tavares2017]          2100      1.008435  \n",
       "log_v|dataset_sd                    1467      1.012255  "
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_parameter_behavior_e_v_rt.csv')\n",
    "\n",
    "if not os.path.isfile(result_file):\n",
    "\n",
    "    model_e = Model(df)\n",
    "    result_e = model_e.fit('log_rt ~ log_v',\n",
    "                           random=['log_v|dataset'],\n",
    "                           categorical=['dataset'],\n",
    "                           samples=n_samples,\n",
    "                           chains=2)\n",
    "\n",
    "    summary_e = result_e.summary(ranefs=True)\n",
    "    summary_e.to_csv(result_file)\n",
    "else:\n",
    "    print('Loading previous results from \"{}\"...'.format(result_file))\n",
    "    summary_e = pd.read_csv(result_file, index_col=0)\n",
    "summary_e"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading previous results from \"results/mixed_effects_models/si2_parameter_behavior_f_gamma_best_chosen.csv\"...\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "      <th>P(&gt;0)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>-5.956513</td>\n",
       "      <td>4.364543</td>\n",
       "      <td>-14.871410</td>\n",
       "      <td>2.138893</td>\n",
       "      <td>1592</td>\n",
       "      <td>1.000831</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>1.396610</td>\n",
       "      <td>4.054223</td>\n",
       "      <td>-6.568707</td>\n",
       "      <td>9.555024</td>\n",
       "      <td>1428</td>\n",
       "      <td>1.001228</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>-1.574253</td>\n",
       "      <td>4.180720</td>\n",
       "      <td>-10.024388</td>\n",
       "      <td>6.637936</td>\n",
       "      <td>1579</td>\n",
       "      <td>1.000950</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>7.815341</td>\n",
       "      <td>4.592465</td>\n",
       "      <td>-0.879668</td>\n",
       "      <td>16.995088</td>\n",
       "      <td>1711</td>\n",
       "      <td>1.000650</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>7.232852</td>\n",
       "      <td>3.171432</td>\n",
       "      <td>1.807807</td>\n",
       "      <td>13.893732</td>\n",
       "      <td>2905</td>\n",
       "      <td>1.000337</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>81.765392</td>\n",
       "      <td>3.981087</td>\n",
       "      <td>73.935827</td>\n",
       "      <td>89.675620</td>\n",
       "      <td>1283</td>\n",
       "      <td>1.001253</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>best_chosen_sd</th>\n",
       "      <td>8.065718</td>\n",
       "      <td>0.552239</td>\n",
       "      <td>7.010017</td>\n",
       "      <td>9.160790</td>\n",
       "      <td>31223</td>\n",
       "      <td>0.999975</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma</th>\n",
       "      <td>-0.538628</td>\n",
       "      <td>0.217739</td>\n",
       "      <td>-0.966284</td>\n",
       "      <td>-0.058808</td>\n",
       "      <td>10852</td>\n",
       "      <td>0.999996</td>\n",
       "      <td>0.0207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[folke2016]</th>\n",
       "      <td>-0.140555</td>\n",
       "      <td>0.237380</td>\n",
       "      <td>-0.686321</td>\n",
       "      <td>0.280260</td>\n",
       "      <td>10451</td>\n",
       "      <td>0.999984</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[krajbich2010]</th>\n",
       "      <td>0.091832</td>\n",
       "      <td>0.230552</td>\n",
       "      <td>-0.378361</td>\n",
       "      <td>0.607520</td>\n",
       "      <td>14686</td>\n",
       "      <td>0.999975</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[krajbich2011]</th>\n",
       "      <td>-0.143651</td>\n",
       "      <td>0.261246</td>\n",
       "      <td>-0.731397</td>\n",
       "      <td>0.316797</td>\n",
       "      <td>12857</td>\n",
       "      <td>0.999985</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[tavares2017]</th>\n",
       "      <td>0.128415</td>\n",
       "      <td>0.273246</td>\n",
       "      <td>-0.369849</td>\n",
       "      <td>0.761378</td>\n",
       "      <td>16275</td>\n",
       "      <td>0.999981</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset_sd</th>\n",
       "      <td>0.319963</td>\n",
       "      <td>0.272080</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>0.858766</td>\n",
       "      <td>9667</td>\n",
       "      <td>0.999988</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  mean        sd  hpd0.95_lower  \\\n",
       "1|dataset[folke2016]         -5.956513  4.364543     -14.871410   \n",
       "1|dataset[krajbich2010]       1.396610  4.054223      -6.568707   \n",
       "1|dataset[krajbich2011]      -1.574253  4.180720     -10.024388   \n",
       "1|dataset[tavares2017]        7.815341  4.592465      -0.879668   \n",
       "1|dataset_sd                  7.232852  3.171432       1.807807   \n",
       "Intercept                    81.765392  3.981087      73.935827   \n",
       "best_chosen_sd                8.065718  0.552239       7.010017   \n",
       "gamma                        -0.538628  0.217739      -0.966284   \n",
       "gamma|dataset[folke2016]     -0.140555  0.237380      -0.686321   \n",
       "gamma|dataset[krajbich2010]   0.091832  0.230552      -0.378361   \n",
       "gamma|dataset[krajbich2011]  -0.143651  0.261246      -0.731397   \n",
       "gamma|dataset[tavares2017]    0.128415  0.273246      -0.369849   \n",
       "gamma|dataset_sd              0.319963  0.272080       0.000010   \n",
       "\n",
       "                             hpd0.95_upper  effective_n  gelman_rubin   P(>0)  \n",
       "1|dataset[folke2016]              2.138893         1592      1.000831     NaN  \n",
       "1|dataset[krajbich2010]           9.555024         1428      1.001228     NaN  \n",
       "1|dataset[krajbich2011]           6.637936         1579      1.000950     NaN  \n",
       "1|dataset[tavares2017]           16.995088         1711      1.000650     NaN  \n",
       "1|dataset_sd                     13.893732         2905      1.000337     NaN  \n",
       "Intercept                        89.675620         1283      1.001253     NaN  \n",
       "best_chosen_sd                    9.160790        31223      0.999975     NaN  \n",
       "gamma                            -0.058808        10852      0.999996  0.0207  \n",
       "gamma|dataset[folke2016]          0.280260        10451      0.999984     NaN  \n",
       "gamma|dataset[krajbich2010]       0.607520        14686      0.999975     NaN  \n",
       "gamma|dataset[krajbich2011]       0.316797        12857      0.999985     NaN  \n",
       "gamma|dataset[tavares2017]        0.761378        16275      0.999981     NaN  \n",
       "gamma|dataset_sd                  0.858766         9667      0.999988     NaN  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_parameter_behavior_f_gamma_best_chosen.csv')\n",
    "\n",
    "if not os.path.isfile(result_file):\n",
    "    model_f = Model(df)\n",
    "    result_f = model_f.fit('best_chosen ~ gamma',\n",
    "                           random=['gamma|dataset'],\n",
    "                           categorical=['dataset'],\n",
    "                           samples=n_samples,\n",
    "                           chains=2)\n",
    "    summary_f = result_f.summary(ranefs=True)\n",
    "    summary_f.loc['gamma', 'P(>0)'] = np.mean(model_f.backend.trace.get_values('gamma') > 0)\n",
    "    summary_f.to_csv(result_file)\n",
    "else:\n",
    "    print('Loading previous results from \"{}\"...'.format(result_file))\n",
    "    summary_f = pd.read_csv(result_file, index_col=0)\n",
    "summary_f"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results loaded from: results/mixed_effects_models/si2_parameter_behavior_f_gamma_best_chosen_posteriorplot.png\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZ0AAAEaCAYAAADKaorCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x13afc2080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_path = os.path.join('results', 'mixed_effects_models', 'si2_parameter_behavior_f_gamma_best_chosen_posteriorplot.png')\n",
    "if not os.path.isfile(plot_path):\n",
    "    plot_posterior(model_f.backend.trace, varnames=['gamma'], ref_val=0, color='lightblue', point_estimate='mode');\n",
    "    plt.savefig(plot_path, dpi=330)\n",
    "else:\n",
    "    print('Results loaded from: {}'.format(plot_path))\n",
    "    show_img(plot_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading previous results from \"results/mixed_effects_models/si2_parameter_behavior_g_gamma_gaze_influence.csv\"...\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>7.838711</td>\n",
       "      <td>5.920029</td>\n",
       "      <td>-2.964167</td>\n",
       "      <td>20.959486</td>\n",
       "      <td>255</td>\n",
       "      <td>1.004774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>-0.917145</td>\n",
       "      <td>5.152000</td>\n",
       "      <td>-10.203787</td>\n",
       "      <td>11.755268</td>\n",
       "      <td>321</td>\n",
       "      <td>1.003224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>-2.807166</td>\n",
       "      <td>5.433637</td>\n",
       "      <td>-14.244069</td>\n",
       "      <td>8.511666</td>\n",
       "      <td>415</td>\n",
       "      <td>1.002254</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>-5.238224</td>\n",
       "      <td>5.682959</td>\n",
       "      <td>-17.446861</td>\n",
       "      <td>5.351413</td>\n",
       "      <td>576</td>\n",
       "      <td>1.002131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>8.207836</td>\n",
       "      <td>4.639112</td>\n",
       "      <td>0.018719</td>\n",
       "      <td>17.421341</td>\n",
       "      <td>236</td>\n",
       "      <td>1.005395</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>12.555314</td>\n",
       "      <td>5.025853</td>\n",
       "      <td>-0.962339</td>\n",
       "      <td>20.862518</td>\n",
       "      <td>272</td>\n",
       "      <td>1.003865</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma</th>\n",
       "      <td>1.290620</td>\n",
       "      <td>0.469369</td>\n",
       "      <td>0.275879</td>\n",
       "      <td>2.213973</td>\n",
       "      <td>182</td>\n",
       "      <td>1.006930</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[folke2016]</th>\n",
       "      <td>-0.462050</td>\n",
       "      <td>0.479151</td>\n",
       "      <td>-1.393693</td>\n",
       "      <td>0.629058</td>\n",
       "      <td>247</td>\n",
       "      <td>1.004655</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[krajbich2010]</th>\n",
       "      <td>0.717301</td>\n",
       "      <td>0.528250</td>\n",
       "      <td>-0.251101</td>\n",
       "      <td>1.838689</td>\n",
       "      <td>195</td>\n",
       "      <td>1.006853</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[krajbich2011]</th>\n",
       "      <td>0.365913</td>\n",
       "      <td>0.528746</td>\n",
       "      <td>-0.684177</td>\n",
       "      <td>1.458000</td>\n",
       "      <td>202</td>\n",
       "      <td>1.006510</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset[tavares2017]</th>\n",
       "      <td>0.122510</td>\n",
       "      <td>0.559095</td>\n",
       "      <td>-0.957370</td>\n",
       "      <td>1.283814</td>\n",
       "      <td>189</td>\n",
       "      <td>1.005757</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gamma|dataset_sd</th>\n",
       "      <td>0.806972</td>\n",
       "      <td>0.448388</td>\n",
       "      <td>0.167155</td>\n",
       "      <td>1.803883</td>\n",
       "      <td>186</td>\n",
       "      <td>1.005816</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gaze_influence_sd</th>\n",
       "      <td>11.003337</td>\n",
       "      <td>0.768340</td>\n",
       "      <td>9.522336</td>\n",
       "      <td>12.496245</td>\n",
       "      <td>2246</td>\n",
       "      <td>1.000941</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  mean        sd  hpd0.95_lower  \\\n",
       "1|dataset[folke2016]          7.838711  5.920029      -2.964167   \n",
       "1|dataset[krajbich2010]      -0.917145  5.152000     -10.203787   \n",
       "1|dataset[krajbich2011]      -2.807166  5.433637     -14.244069   \n",
       "1|dataset[tavares2017]       -5.238224  5.682959     -17.446861   \n",
       "1|dataset_sd                  8.207836  4.639112       0.018719   \n",
       "Intercept                    12.555314  5.025853      -0.962339   \n",
       "gamma                         1.290620  0.469369       0.275879   \n",
       "gamma|dataset[folke2016]     -0.462050  0.479151      -1.393693   \n",
       "gamma|dataset[krajbich2010]   0.717301  0.528250      -0.251101   \n",
       "gamma|dataset[krajbich2011]   0.365913  0.528746      -0.684177   \n",
       "gamma|dataset[tavares2017]    0.122510  0.559095      -0.957370   \n",
       "gamma|dataset_sd              0.806972  0.448388       0.167155   \n",
       "gaze_influence_sd            11.003337  0.768340       9.522336   \n",
       "\n",
       "                             hpd0.95_upper  effective_n  gelman_rubin  \n",
       "1|dataset[folke2016]             20.959486          255      1.004774  \n",
       "1|dataset[krajbich2010]          11.755268          321      1.003224  \n",
       "1|dataset[krajbich2011]           8.511666          415      1.002254  \n",
       "1|dataset[tavares2017]            5.351413          576      1.002131  \n",
       "1|dataset_sd                     17.421341          236      1.005395  \n",
       "Intercept                        20.862518          272      1.003865  \n",
       "gamma                             2.213973          182      1.006930  \n",
       "gamma|dataset[folke2016]          0.629058          247      1.004655  \n",
       "gamma|dataset[krajbich2010]       1.838689          195      1.006853  \n",
       "gamma|dataset[krajbich2011]       1.458000          202      1.006510  \n",
       "gamma|dataset[tavares2017]        1.283814          189      1.005757  \n",
       "gamma|dataset_sd                  1.803883          186      1.005816  \n",
       "gaze_influence_sd                12.496245         2246      1.000941  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_parameter_behavior_g_gamma_gaze_influence.csv')\n",
    "\n",
    "if not os.path.isfile(result_file):\n",
    "    model_g = Model(df)\n",
    "    result_g = model_g.fit('gaze_influence ~ gamma',\n",
    "                           random=['gamma|dataset'],\n",
    "                           categorical=['dataset'],\n",
    "                           samples=n_samples,\n",
    "                           chains=2)\n",
    "    summary_g = result_g.summary(ranefs=True)\n",
    "    summary_g.to_csv(result_file)\n",
    "else:\n",
    "    print('Loading previous results from \"{}\"...'.format(result_file))\n",
    "    summary_g = pd.read_csv(result_file, index_col=0)\n",
    "summary_g"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Figure SI 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def figure_si_2(waic, \n",
    "                pred_multiplicative_summary, pred_additive_summary, \n",
    "                estimates, odd_summary, summary_e, summary_f, summary_g, \n",
    "                figsize=cm2inch(18, 18)):\n",
    "    \n",
    "    fig = plt.figure(figsize=figsize)\n",
    "    \n",
    "    ax0 = plt.subplot2grid((3, 3), (0, 0), colspan=3)\n",
    "    ax1 = plt.subplot2grid((3, 3), (1, 0))\n",
    "    ax2 = plt.subplot2grid((3, 3), (1, 1))\n",
    "    ax3 = plt.subplot2grid((3, 3), (1, 2))\n",
    "    ax4 = plt.subplot2grid((3, 3), (2, 0))\n",
    "    ax5 = plt.subplot2grid((3, 3), (2, 1))\n",
    "    ax6 = plt.subplot2grid((3, 3), (2, 2))\n",
    "    \n",
    "    figure_si_2a(waic, ylim=(-5, 5), ax=ax0)\n",
    "    figure_si_2bcd(pred_multiplicative_summary, pred_additive_summary, axs=np.array([ax1, ax2, ax3]), plot_legend=False)\n",
    "    figure_si_2efg(estimates, odd_summary, axs=np.array([ax4, ax5, ax6]), plot_legend=False)\n",
    "    \n",
    "    add_regression_line(ax4, summary_e.loc['Intercept', 'mean'], summary_e.loc['log_v', 'mean'], linewidth=1)\n",
    "    add_regression_line(ax5, summary_f.loc['Intercept', 'mean'] / 100, summary_f.loc['gamma', 'mean'] / 100, linewidth=1)\n",
    "    add_regression_line(ax6, summary_g.loc['Intercept', 'mean'] / 100, summary_g.loc['gamma', 'mean'] / 100, linewidth=1)\n",
    "    \n",
    "    fig.tight_layout()\n",
    "    \n",
    "    return fig, [ax0, ax1, ax2, ax3, ax4, ax5, ax6]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgwAAAINCAYAAABS2ieqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTf3V1qD/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x152a0a0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "figure_si_2(waic,\n",
    "            pred_multiplicative_summary, pred_additive_summary, \n",
    "            estimates, odd_summary, summary_e, summary_f, summary_g,\n",
    "            figsize=cm2inch(18.8, 18.8))\n",
    "plt.savefig('results/figures/si_figure_2_multiplicative_vs_additive.png', dpi=330, bbox_inches='tight')\n",
    "plt.savefig('results/figures/si_figure_2_multiplicative_vs_additive.pdf', dpi=330, bbox_inches='tight')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bias Anlaysis: Multiplicative vs additive (b - d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>best_chosen</th>\n",
       "      <th>dataset</th>\n",
       "      <th>gaze_influence</th>\n",
       "      <th>is_pred</th>\n",
       "      <th>mean_rt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.904889</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.077317</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1434.192000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.640000</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.216067</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1443.249796</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.836000</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.007773</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4486.572000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.656818</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.128273</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1966.197273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.744800</td>\n",
       "      <td>krajbich2010</td>\n",
       "      <td>0.387967</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1650.926400</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   best_chosen       dataset  gaze_influence  is_pred      mean_rt\n",
       "0     0.904889  krajbich2010        0.077317      0.0  1434.192000\n",
       "1     0.640000  krajbich2010        0.216067      0.0  1443.249796\n",
       "2     0.836000  krajbich2010        0.007773      0.0  4486.572000\n",
       "3     0.656818  krajbich2010        0.128273      0.0  1966.197273\n",
       "4     0.744800  krajbich2010        0.387967      0.0  1650.926400"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Prepare data for mixed model\n",
    "mean_rt = np.concatenate([pred_multiplicative_summary['rt']['mean'].values,\n",
    "                          pred_additive_summary['rt']['mean'].values],\n",
    "                         axis=0)\n",
    "best_chosen = np.concatenate([pred_multiplicative_summary['best_chosen']['mean'].values,\n",
    "                              pred_additive_summary['best_chosen']['mean'].values],\n",
    "                             axis=0)\n",
    "gaze_influence = np.concatenate([pred_multiplicative_summary['gaze_influence']['Unnamed: 10_level_1'].values,\n",
    "                                 pred_additive_summary['gaze_influence']['Unnamed: 10_level_1'].values],\n",
    "                                axis=0)\n",
    "\n",
    "is_pred = np.concatenate([np.zeros(len(pred_multiplicative_summary)),\n",
    "                          np.ones(len(pred_additive_summary))],\n",
    "                         axis=0)\n",
    "dataset = np.concatenate([pred_multiplicative_summary['dataset']['Unnamed: 11_level_1'].values,\n",
    "                          pred_additive_summary['dataset']['Unnamed: 11_level_1'].values],\n",
    "                         axis=0)\n",
    "\n",
    "df = pd.DataFrame(dict(mean_rt=mean_rt,\n",
    "                       best_chosen=best_chosen,\n",
    "                       gaze_influence=gaze_influence,\n",
    "                       is_pred=is_pred,\n",
    "                       dataset=dataset))\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4388: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  object.__getattribute__(self, name)\n",
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4389: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  return object.__setattr__(self, name, value)\n",
      "Average Loss = 2,002.2: 100%|██████████| 50000/50000 [00:24<00:00, 2049.82it/s]\n",
      "100%|██████████| 20500/20500 [06:41<00:00, 51.02it/s]\n",
      "100%|██████████| 20500/20500 [06:37<00:00, 51.57it/s]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>2411.939624</td>\n",
       "      <td>783.682426</td>\n",
       "      <td>945.532115</td>\n",
       "      <td>4083.413845</td>\n",
       "      <td>1037</td>\n",
       "      <td>1.003806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>-754.656982</td>\n",
       "      <td>780.124260</td>\n",
       "      <td>-2332.521320</td>\n",
       "      <td>782.723822</td>\n",
       "      <td>1028</td>\n",
       "      <td>1.004018</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>-472.190196</td>\n",
       "      <td>780.180037</td>\n",
       "      <td>-1994.365735</td>\n",
       "      <td>1134.551428</td>\n",
       "      <td>1034</td>\n",
       "      <td>1.003866</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>-974.565860</td>\n",
       "      <td>781.390434</td>\n",
       "      <td>-2540.822790</td>\n",
       "      <td>578.008808</td>\n",
       "      <td>1044</td>\n",
       "      <td>1.003929</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>1603.209883</td>\n",
       "      <td>510.863162</td>\n",
       "      <td>768.785500</td>\n",
       "      <td>2627.723891</td>\n",
       "      <td>5484</td>\n",
       "      <td>0.999998</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>2885.032930</td>\n",
       "      <td>773.930468</td>\n",
       "      <td>1319.723919</td>\n",
       "      <td>4411.363359</td>\n",
       "      <td>994</td>\n",
       "      <td>1.004175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred</th>\n",
       "      <td>54.091147</td>\n",
       "      <td>188.102917</td>\n",
       "      <td>-321.004097</td>\n",
       "      <td>446.042543</td>\n",
       "      <td>1728</td>\n",
       "      <td>1.001060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[folke2016]</th>\n",
       "      <td>19.887457</td>\n",
       "      <td>195.527131</td>\n",
       "      <td>-406.321185</td>\n",
       "      <td>449.034240</td>\n",
       "      <td>3814</td>\n",
       "      <td>1.000408</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2010]</th>\n",
       "      <td>-13.975746</td>\n",
       "      <td>185.285549</td>\n",
       "      <td>-433.450621</td>\n",
       "      <td>384.183365</td>\n",
       "      <td>2665</td>\n",
       "      <td>1.000679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2011]</th>\n",
       "      <td>13.146100</td>\n",
       "      <td>192.480963</td>\n",
       "      <td>-406.446400</td>\n",
       "      <td>441.020989</td>\n",
       "      <td>3185</td>\n",
       "      <td>1.000508</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[tavares2017]</th>\n",
       "      <td>-25.371762</td>\n",
       "      <td>193.975654</td>\n",
       "      <td>-475.582511</td>\n",
       "      <td>379.285588</td>\n",
       "      <td>3730</td>\n",
       "      <td>1.000560</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset_sd</th>\n",
       "      <td>225.162695</td>\n",
       "      <td>231.819545</td>\n",
       "      <td>0.020617</td>\n",
       "      <td>682.846892</td>\n",
       "      <td>3382</td>\n",
       "      <td>1.000202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean_rt_sd</th>\n",
       "      <td>1043.835048</td>\n",
       "      <td>48.963561</td>\n",
       "      <td>948.714566</td>\n",
       "      <td>1140.473810</td>\n",
       "      <td>40000</td>\n",
       "      <td>0.999976</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                      mean          sd  hpd0.95_lower  \\\n",
       "1|dataset[folke2016]           2411.939624  783.682426     945.532115   \n",
       "1|dataset[krajbich2010]        -754.656982  780.124260   -2332.521320   \n",
       "1|dataset[krajbich2011]        -472.190196  780.180037   -1994.365735   \n",
       "1|dataset[tavares2017]         -974.565860  781.390434   -2540.822790   \n",
       "1|dataset_sd                   1603.209883  510.863162     768.785500   \n",
       "Intercept                      2885.032930  773.930468    1319.723919   \n",
       "is_pred                          54.091147  188.102917    -321.004097   \n",
       "is_pred|dataset[folke2016]       19.887457  195.527131    -406.321185   \n",
       "is_pred|dataset[krajbich2010]   -13.975746  185.285549    -433.450621   \n",
       "is_pred|dataset[krajbich2011]    13.146100  192.480963    -406.446400   \n",
       "is_pred|dataset[tavares2017]    -25.371762  193.975654    -475.582511   \n",
       "is_pred|dataset_sd              225.162695  231.819545       0.020617   \n",
       "mean_rt_sd                     1043.835048   48.963561     948.714566   \n",
       "\n",
       "                               hpd0.95_upper  effective_n  gelman_rubin  \n",
       "1|dataset[folke2016]             4083.413845         1037      1.003806  \n",
       "1|dataset[krajbich2010]           782.723822         1028      1.004018  \n",
       "1|dataset[krajbich2011]          1134.551428         1034      1.003866  \n",
       "1|dataset[tavares2017]            578.008808         1044      1.003929  \n",
       "1|dataset_sd                     2627.723891         5484      0.999998  \n",
       "Intercept                        4411.363359          994      1.004175  \n",
       "is_pred                           446.042543         1728      1.001060  \n",
       "is_pred|dataset[folke2016]        449.034240         3814      1.000408  \n",
       "is_pred|dataset[krajbich2010]     384.183365         2665      1.000679  \n",
       "is_pred|dataset[krajbich2011]     441.020989         3185      1.000508  \n",
       "is_pred|dataset[tavares2017]      379.285588         3730      1.000560  \n",
       "is_pred|dataset_sd                682.846892         3382      1.000202  \n",
       "mean_rt_sd                       1140.473810        40000      0.999976  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Response time\n",
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_multiplicative_vs_additive_b_bias_analysis_mean_rt.csv')\n",
    "if not os.path.isfile(result_file):\n",
    "    model = Model(df)\n",
    "    results_mean_rt = model.fit('mean_rt ~ is_pred',\n",
    "                                random=['is_pred|dataset'],\n",
    "                                categorical=['dataset'],\n",
    "                                samples=n_samples)\n",
    "    bias_analysis_rt = results_mean_rt.summary(ranefs=True)\n",
    "    bias_analysis_rt.to_csv(result_file)\n",
    "else:\n",
    "    bias_analysis_rt = pd.read_csv(result_file)\n",
    "bias_analysis_rt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4388: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  object.__getattribute__(self, name)\n",
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4389: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  return object.__setattr__(self, name, value)\n",
      "Average Loss = -184.86:  55%|█████▍    | 27371/50000 [00:15<00:12, 1753.35it/s]\n",
      "100%|██████████| 20500/20500 [02:18<00:00, 147.73it/s]\n",
      "100%|██████████| 20500/20500 [02:20<00:00, 146.10it/s]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>-0.120035</td>\n",
       "      <td>0.060581</td>\n",
       "      <td>-2.437992e-01</td>\n",
       "      <td>-0.002041</td>\n",
       "      <td>1463</td>\n",
       "      <td>1.000732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>0.066649</td>\n",
       "      <td>0.060176</td>\n",
       "      <td>-5.675721e-02</td>\n",
       "      <td>0.183044</td>\n",
       "      <td>1491</td>\n",
       "      <td>1.000825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>-0.042117</td>\n",
       "      <td>0.060181</td>\n",
       "      <td>-1.666782e-01</td>\n",
       "      <td>0.073122</td>\n",
       "      <td>1434</td>\n",
       "      <td>1.000832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>0.100450</td>\n",
       "      <td>0.060592</td>\n",
       "      <td>-2.251710e-02</td>\n",
       "      <td>0.217242</td>\n",
       "      <td>1415</td>\n",
       "      <td>1.000883</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>0.112360</td>\n",
       "      <td>0.039678</td>\n",
       "      <td>4.858309e-02</td>\n",
       "      <td>0.193332</td>\n",
       "      <td>4040</td>\n",
       "      <td>1.000420</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>0.744138</td>\n",
       "      <td>0.059464</td>\n",
       "      <td>6.266083e-01</td>\n",
       "      <td>0.864358</td>\n",
       "      <td>1381</td>\n",
       "      <td>1.000790</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>best_chosen_sd</th>\n",
       "      <td>0.100208</td>\n",
       "      <td>0.004697</td>\n",
       "      <td>9.089373e-02</td>\n",
       "      <td>0.109263</td>\n",
       "      <td>31879</td>\n",
       "      <td>1.000028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred</th>\n",
       "      <td>-0.001118</td>\n",
       "      <td>0.020210</td>\n",
       "      <td>-4.017085e-02</td>\n",
       "      <td>0.039644</td>\n",
       "      <td>5135</td>\n",
       "      <td>0.999992</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[folke2016]</th>\n",
       "      <td>-0.003717</td>\n",
       "      <td>0.020709</td>\n",
       "      <td>-4.976389e-02</td>\n",
       "      <td>0.038933</td>\n",
       "      <td>6722</td>\n",
       "      <td>1.000263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2010]</th>\n",
       "      <td>0.000312</td>\n",
       "      <td>0.019871</td>\n",
       "      <td>-4.453433e-02</td>\n",
       "      <td>0.040837</td>\n",
       "      <td>4895</td>\n",
       "      <td>1.000034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2011]</th>\n",
       "      <td>0.000564</td>\n",
       "      <td>0.020436</td>\n",
       "      <td>-4.289747e-02</td>\n",
       "      <td>0.044208</td>\n",
       "      <td>5262</td>\n",
       "      <td>0.999980</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[tavares2017]</th>\n",
       "      <td>0.002362</td>\n",
       "      <td>0.020549</td>\n",
       "      <td>-4.007905e-02</td>\n",
       "      <td>0.049116</td>\n",
       "      <td>6145</td>\n",
       "      <td>0.999993</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset_sd</th>\n",
       "      <td>0.022912</td>\n",
       "      <td>0.025235</td>\n",
       "      <td>8.474200e-07</td>\n",
       "      <td>0.071340</td>\n",
       "      <td>3217</td>\n",
       "      <td>1.000002</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   mean        sd  hpd0.95_lower  \\\n",
       "1|dataset[folke2016]          -0.120035  0.060581  -2.437992e-01   \n",
       "1|dataset[krajbich2010]        0.066649  0.060176  -5.675721e-02   \n",
       "1|dataset[krajbich2011]       -0.042117  0.060181  -1.666782e-01   \n",
       "1|dataset[tavares2017]         0.100450  0.060592  -2.251710e-02   \n",
       "1|dataset_sd                   0.112360  0.039678   4.858309e-02   \n",
       "Intercept                      0.744138  0.059464   6.266083e-01   \n",
       "best_chosen_sd                 0.100208  0.004697   9.089373e-02   \n",
       "is_pred                       -0.001118  0.020210  -4.017085e-02   \n",
       "is_pred|dataset[folke2016]    -0.003717  0.020709  -4.976389e-02   \n",
       "is_pred|dataset[krajbich2010]  0.000312  0.019871  -4.453433e-02   \n",
       "is_pred|dataset[krajbich2011]  0.000564  0.020436  -4.289747e-02   \n",
       "is_pred|dataset[tavares2017]   0.002362  0.020549  -4.007905e-02   \n",
       "is_pred|dataset_sd             0.022912  0.025235   8.474200e-07   \n",
       "\n",
       "                               hpd0.95_upper  effective_n  gelman_rubin  \n",
       "1|dataset[folke2016]               -0.002041         1463      1.000732  \n",
       "1|dataset[krajbich2010]             0.183044         1491      1.000825  \n",
       "1|dataset[krajbich2011]             0.073122         1434      1.000832  \n",
       "1|dataset[tavares2017]              0.217242         1415      1.000883  \n",
       "1|dataset_sd                        0.193332         4040      1.000420  \n",
       "Intercept                           0.864358         1381      1.000790  \n",
       "best_chosen_sd                      0.109263        31879      1.000028  \n",
       "is_pred                             0.039644         5135      0.999992  \n",
       "is_pred|dataset[folke2016]          0.038933         6722      1.000263  \n",
       "is_pred|dataset[krajbich2010]       0.040837         4895      1.000034  \n",
       "is_pred|dataset[krajbich2011]       0.044208         5262      0.999980  \n",
       "is_pred|dataset[tavares2017]        0.049116         6145      0.999993  \n",
       "is_pred|dataset_sd                  0.071340         3217      1.000002  "
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# P(choose best)\n",
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_multiplicative_vs_additive_c_bias_analysis_p_choose_best.csv')\n",
    "if not os.path.isfile(result_file):\n",
    "    model = Model(df)\n",
    "    results_best_chosen = model.fit('best_chosen ~ is_pred',\n",
    "                                    random=['is_pred|dataset'],\n",
    "                                    categorical=['dataset'],\n",
    "                                    samples=n_samples)\n",
    "    bias_analysis_best_chosen = results_best_chosen.summary(ranefs=True)\n",
    "    bias_analysis_best_chosen.to_csv(result_file)\n",
    "else:\n",
    "    bias_analysis_best_chosen = pd.read_csv(result_file)\n",
    "bias_analysis_best_chosen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4388: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  object.__getattribute__(self, name)\n",
      "/Users/armin/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py:4389: FutureWarning: Attribute 'is_copy' is deprecated and will be removed in a future version.\n",
      "  return object.__setattr__(self, name, value)\n",
      "Average Loss = -127.74:  47%|████▋     | 23650/50000 [00:14<00:15, 1674.55it/s]\n",
      "100%|██████████| 20500/20500 [01:45<00:00, 194.21it/s]\n",
      "100%|██████████| 20500/20500 [01:52<00:00, 182.65it/s]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>sd</th>\n",
       "      <th>hpd0.95_lower</th>\n",
       "      <th>hpd0.95_upper</th>\n",
       "      <th>effective_n</th>\n",
       "      <th>gelman_rubin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1|dataset[folke2016]</th>\n",
       "      <td>0.074984</td>\n",
       "      <td>0.042236</td>\n",
       "      <td>-0.005754</td>\n",
       "      <td>0.164354</td>\n",
       "      <td>6441</td>\n",
       "      <td>0.999975</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2010]</th>\n",
       "      <td>-0.051544</td>\n",
       "      <td>0.040483</td>\n",
       "      <td>-0.135949</td>\n",
       "      <td>0.030096</td>\n",
       "      <td>5836</td>\n",
       "      <td>1.000012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[krajbich2011]</th>\n",
       "      <td>0.001115</td>\n",
       "      <td>0.040879</td>\n",
       "      <td>-0.084181</td>\n",
       "      <td>0.082608</td>\n",
       "      <td>6073</td>\n",
       "      <td>0.999975</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset[tavares2017]</th>\n",
       "      <td>-0.018472</td>\n",
       "      <td>0.041482</td>\n",
       "      <td>-0.107063</td>\n",
       "      <td>0.062718</td>\n",
       "      <td>6055</td>\n",
       "      <td>0.999989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1|dataset_sd</th>\n",
       "      <td>0.074467</td>\n",
       "      <td>0.035068</td>\n",
       "      <td>0.021448</td>\n",
       "      <td>0.144084</td>\n",
       "      <td>5703</td>\n",
       "      <td>0.999979</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Intercept</th>\n",
       "      <td>0.217709</td>\n",
       "      <td>0.039171</td>\n",
       "      <td>0.138780</td>\n",
       "      <td>0.301128</td>\n",
       "      <td>5342</td>\n",
       "      <td>0.999978</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gaze_influence_sd</th>\n",
       "      <td>0.129316</td>\n",
       "      <td>0.005046</td>\n",
       "      <td>0.120500</td>\n",
       "      <td>0.138645</td>\n",
       "      <td>17965</td>\n",
       "      <td>1.000318</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred</th>\n",
       "      <td>-0.036538</td>\n",
       "      <td>0.027552</td>\n",
       "      <td>-0.095454</td>\n",
       "      <td>0.015632</td>\n",
       "      <td>4043</td>\n",
       "      <td>1.000270</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[folke2016]</th>\n",
       "      <td>-0.005839</td>\n",
       "      <td>0.029356</td>\n",
       "      <td>-0.073901</td>\n",
       "      <td>0.051738</td>\n",
       "      <td>5214</td>\n",
       "      <td>1.000248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2010]</th>\n",
       "      <td>-0.003809</td>\n",
       "      <td>0.027802</td>\n",
       "      <td>-0.065703</td>\n",
       "      <td>0.053565</td>\n",
       "      <td>4615</td>\n",
       "      <td>1.000388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[krajbich2011]</th>\n",
       "      <td>-0.004018</td>\n",
       "      <td>0.028423</td>\n",
       "      <td>-0.068171</td>\n",
       "      <td>0.052262</td>\n",
       "      <td>4100</td>\n",
       "      <td>1.000188</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset[tavares2017]</th>\n",
       "      <td>0.012761</td>\n",
       "      <td>0.030324</td>\n",
       "      <td>-0.042232</td>\n",
       "      <td>0.082714</td>\n",
       "      <td>4804</td>\n",
       "      <td>1.000177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>is_pred|dataset_sd</th>\n",
       "      <td>0.033624</td>\n",
       "      <td>0.035173</td>\n",
       "      <td>0.000005</td>\n",
       "      <td>0.101348</td>\n",
       "      <td>3656</td>\n",
       "      <td>1.000185</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   mean        sd  hpd0.95_lower  \\\n",
       "1|dataset[folke2016]           0.074984  0.042236      -0.005754   \n",
       "1|dataset[krajbich2010]       -0.051544  0.040483      -0.135949   \n",
       "1|dataset[krajbich2011]        0.001115  0.040879      -0.084181   \n",
       "1|dataset[tavares2017]        -0.018472  0.041482      -0.107063   \n",
       "1|dataset_sd                   0.074467  0.035068       0.021448   \n",
       "Intercept                      0.217709  0.039171       0.138780   \n",
       "gaze_influence_sd              0.129316  0.005046       0.120500   \n",
       "is_pred                       -0.036538  0.027552      -0.095454   \n",
       "is_pred|dataset[folke2016]    -0.005839  0.029356      -0.073901   \n",
       "is_pred|dataset[krajbich2010] -0.003809  0.027802      -0.065703   \n",
       "is_pred|dataset[krajbich2011] -0.004018  0.028423      -0.068171   \n",
       "is_pred|dataset[tavares2017]   0.012761  0.030324      -0.042232   \n",
       "is_pred|dataset_sd             0.033624  0.035173       0.000005   \n",
       "\n",
       "                               hpd0.95_upper  effective_n  gelman_rubin  \n",
       "1|dataset[folke2016]                0.164354         6441      0.999975  \n",
       "1|dataset[krajbich2010]             0.030096         5836      1.000012  \n",
       "1|dataset[krajbich2011]             0.082608         6073      0.999975  \n",
       "1|dataset[tavares2017]              0.062718         6055      0.999989  \n",
       "1|dataset_sd                        0.144084         5703      0.999979  \n",
       "Intercept                           0.301128         5342      0.999978  \n",
       "gaze_influence_sd                   0.138645        17965      1.000318  \n",
       "is_pred                             0.015632         4043      1.000270  \n",
       "is_pred|dataset[folke2016]          0.051738         5214      1.000248  \n",
       "is_pred|dataset[krajbich2010]       0.053565         4615      1.000388  \n",
       "is_pred|dataset[krajbich2011]       0.052262         4100      1.000188  \n",
       "is_pred|dataset[tavares2017]        0.082714         4804      1.000177  \n",
       "is_pred|dataset_sd                  0.101348         3656      1.000185  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Gaze influence on P(choose | value)\n",
    "result_file = os.path.join('results', 'mixed_effects_models', 'si2_multiplicative_vs_additive_d_bias_analysis_gaze_influence.csv')\n",
    "if not os.path.isfile(result_file):\n",
    "    model = Model(df)\n",
    "    results_gaze_influence = model.fit('gaze_influence ~ is_pred',\n",
    "                                       random=['is_pred|dataset'],\n",
    "                                       categorical=['dataset'],\n",
    "                                       samples=n_samples)\n",
    "    bias_analysis_gaze_influence = results_gaze_influence.summary(ranefs=True)\n",
    "    bias_analysis_gaze_influence.to_csv(result_file)\n",
    "else:\n",
    "    bias_analysis_gaze_influence = pd.read_csv(result_file)\n",
    "bias_analysis_gaze_influence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  },
  "toc": {
   "colors": {
    "hover_highlight": "#DAA520",
    "running_highlight": "#FF0000",
    "selected_highlight": "#FFD700"
   },
   "moveMenuLeft": true,
   "nav_menu": {
    "height": "156px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": false,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false,
   "widenNotebook": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}