DEMO: practical IAT

- scoring script included (good) but requires command line (bad).

psychopy/demos/builder/practical IAT/All.csv

@@ -0,0 +1,25 @@
+word,condition,key,color
+Creative,creative,e,skyblue
+Novel,creative,e,skyblue
+Inventive,creative,e,skyblue
+Newfangled,creative,e,skyblue
+Original,creative,e,skyblue
+Different,creative,e,skyblue
+Practical,practical,i,skyblue
+Applied,practical,i,skyblue
+Useful,practical,i,skyblue
+Functional,practical,i,skyblue
+Efficient,practical,i,skyblue
+Constructive,practical,i,skyblue
+Rainbow,good,i,limegreen
+Cake,good,i,limegreen
+Sunshine,good,i,limegreen
+Laughter,good,i,limegreen
+Peace,good,i,limegreen
+Heaven,good,i,limegreen
+Agony,bad,e,limegreen
+Hell,bad,e,limegreen
+Ugly,bad,e,limegreen
+Vomit,bad,e,limegreen
+Poison,bad,e,limegreen
+Rotten,bad,e,limegreen

psychopy/demos/builder/practical IAT/All_rev.csv

@@ -0,0 +1,25 @@
+word,condition,key,color
+Creative,creative,i,skyblue
+Novel,creative,i,skyblue
+Inventive,creative,i,skyblue
+Newfangled,creative,i,skyblue
+Original,creative,i,skyblue
+Different,creative,i,skyblue
+Practical,practical,e,skyblue
+Applied,practical,e,skyblue
+Useful,practical,e,skyblue
+Functional,practical,e,skyblue
+Efficient,practical,e,skyblue
+Constructive,practical,e,skyblue
+Rainbow,good,i,limegreen
+Cake,good,i,limegreen
+Sunshine,good,i,limegreen
+Laughter,good,i,limegreen
+Peace,good,i,limegreen
+Heaven,good,i,limegreen
+Agony,bad,e,limegreen
+Hell,bad,e,limegreen
+Ugly,bad,e,limegreen
+Vomit,bad,e,limegreen
+Poison,bad,e,limegreen
+Rotten,bad,e,limegreen

psychopy/demos/builder/practical IAT/Creat_Prac.csv

@@ -0,0 +1,13 @@
+word,condition,key,color
+Creative,creative,e,skyblue
+Novel,creative,e,skyblue
+Inventive,creative,e,skyblue
+Newfangled,creative,e,skyblue
+Original,creative,e,skyblue
+Different,creative,e,skyblue
+Practical,practical,i,skyblue
+Applied,practical,i,skyblue
+Useful,practical,i,skyblue
+Functional,practical,i,skyblue
+Efficient,practical,i,skyblue
+Constructive,practical,i,skyblue

psychopy/demos/builder/practical IAT/Creat_Prac_rev.csv

@@ -0,0 +1,13 @@
+word,condition,key,color
+Creative,creative,i,skyblue
+Novel,creative,i,skyblue
+Inventive,creative,i,skyblue
+Newfangled,creative,i,skyblue
+Original,creative,i,skyblue
+Different,creative,i,skyblue
+Practical,practical,e,skyblue
+Applied,practical,e,skyblue
+Useful,practical,e,skyblue
+Functional,practical,e,skyblue
+Efficient,practical,e,skyblue
+Constructive,practical,e,skyblue

psychopy/demos/builder/practical IAT/Good_Bad.csv

@@ -0,0 +1,13 @@
+word,condition,key,color
+Rainbow,good,i,limegreen
+Cake,good,i,limegreen
+Sunshine,good,i,limegreen
+Laughter,good,i,limegreen
+Peace,good,i,limegreen
+Heaven,good,i,limegreen
+Agony,bad,e,limegreen
+Hell,bad,e,limegreen
+Ugly,bad,e,limegreen
+Vomit,bad,e,limegreen
+Poison,bad,e,limegreen
+Rotten,bad,e,limegreen

psychopy/demos/builder/practical IAT/README.TXT

@@ -0,0 +1,25 @@
+        A practical IAT in PsychoPy: Creative-Practical x Good-Bad
+
+Copyright (C) Jeremy R. Gray, 2014. Released under the terms of the GPLv3.
+Contact jrgray@gmail.com
+
+Authors: Jeremy R. Gray, Nathaniel Pasmanter
+
+Task modeled after that reported by JS Mueller, et al.,‎2011, Psych Science.
+
+To obtain a bias score, use the script provided in the "Practical IAT" directory:
+% python scoreIAT.py data/data_file_name.csv
+data_file_name.csv 0.081
+
+A positive D value (e.g., 0.081) indicates a bias in favor of "practical good / creative bad".
+
+Uses scoring procedure outlined in Nosek et al. (2007), developed by Greenwald Nosek & Banaji (2003), except that RT on incorrect responses do not contribute to average RT (and there's no RT penalty for incorrect responses).
+
+Checks for following exclusionary criteria:
+ - 25% of responses too fast in any one of the critical blocks
+ - going too fast (<300 ms) on more than 10% of responses across all critical blocks
+ - 35% too fast in any one of the practice blocks
+ - making more than 30% erroneous responses across the critical blocks
+ - 40% errors in any one of the critical blocks
+ - 40% errors across all of the practice blocks
+ - 50% errors in any one of the practice blocks

psychopy/demos/builder/practical IAT/scoreIAT.py

@@ -0,0 +1,165 @@
+#!/usr/bin/env python2
+
+"""Scoring script for MSU-PsychoPy version of IAT task.
+
+Authors: Jeremy R. Gray & Nate Pasmanter, 2013
+"""
+
+from __future__ import division
+import pandas as pd
+import glob, os, sys
+
+def scoreIAT(csvfile, write_file=False):
+    """Input = csv file; output = D score (or explanation why data are bad).
+
+    Expects column headers of the form Response_#.corr, and Response_#.rt,
+    where # is 1-7, for IAT block number.
+
+    Scoring is mostly per GNB 2003:
+      Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and
+      using the implicit association test: I. An improved scoring algorithm.
+      Journal of Personality and Social Psychology, 85, 197-216.
+    Following Amodio, incorrect responses do not contribute to RT; there's no
+    RT penalty for this reason. For computing SDs, unbiased SD is used (N-1).
+
+    If write_file=True, will save score to a file 'Scored_' + csvfile.
+
+    A positive D value from this script indicates a bias in favor of
+    "creative bad / practical good". I.e., if RT in blocks with creative+good is
+    longer than RT in blocks with creative+bad, people are more conflicted or
+    hesitant about creative+good.
+    The way the task is set up, when side == 1, creative and bad are paired first.
+    If side == -1, the opposite is true. This pairing is handled by the scoring
+    script.
+    """
+    # ------------ Thresholds for excluding trials or subjects: ------------
+    rt_FAST = 0.300
+    rt_FASTms = int(1000 * rt_FAST)  # 300ms
+    rt_SLOW = 10.
+    correct = 1
+    incorrect = 0
+    # GNB 2003 thesholds for why subject should be excluded:
+    warn = u''
+    threshold = {'ac_prac_blk': 0.50,
+                 'ac_prac_all': 0.60, 'rt_prac_all': 0.35,
+                 'ac_task_blk': 0.60, 'rt_task_blk': 0.25,
+                 'ac_task_all': 0.70, 'rt_task_all': 0.10 }
+
+    # ------------ Read dataframe (df) from .csv file and parse: ------------
+    df = pd.read_csv(csvfile)
+
+    # accuracy; mean --> proportion correct
+    prac_ac = [df.loc[:, 'Response_1.corr'].dropna(),
+               df.loc[:, 'Response_2.corr'].dropna(),
+               df.loc[:, 'Response_5.corr'].dropna()]
+    task_ac = [df.loc[:, 'Response_3.corr'].dropna(),
+               df.loc[:, 'Response_4.corr'].dropna(),
+               df.loc[:, 'Response_6.corr'].dropna(),
+               df.loc[:, 'Response_7.corr'].dropna()]
+    # response time in seconds
+    prac_rt = [df.loc[:, 'Response_1.rt'].dropna(),
+               df.loc[:, 'Response_2.rt'].dropna(),
+               df.loc[:, 'Response_5.rt'].dropna()]
+    task_rt = [df.loc[:, 'Response_3.rt'].dropna(),
+               df.loc[:, 'Response_4.rt'].dropna(),
+               df.loc[:, 'Response_6.rt'].dropna(),
+               df.loc[:, 'Response_7.rt'].dropna()]
+
+    assert len(task_ac[0]) == len(task_ac[2]) == len(task_rt[0])  # block 3, 6
+    assert len(task_ac[1]) == len(task_ac[3]) == len(task_rt[1])  # block 4, 7
+    assert len(task_rt[0]) == len(task_rt[2]) > 1  # require 2+ items in 3, 6
+    assert len(task_rt[1]) == len(task_rt[3]) > 1  # equire 2+ items in 4, 7
+    assert all([all(task_ac[i].isin([correct, incorrect])) for i in range(4)])
+    assert all([all(task_rt[i] > 0) for i in range(4)])  # require positive RTs
+
+    # counterbalanced IAT screen side: +1 or -1; used in calc of D
+    side = df.loc[0, 'side']
+    assert side in [-1, 1]
+
+    # ------------ Check participant exclusion thresholds ------------
+    # check proportion-too-fast in each task block:
+    for i, rt in enumerate(task_rt):
+        prop_too_fast = len(rt[(rt < rt_FAST)]) / len(rt)
+        if prop_too_fast > threshold['rt_task_blk']:
+            pct = 100 * prop_too_fast
+            warn += "%.0f%% trials with RT < %dms in task block #%d\n" % (
+                pct, rt_FASTms, (3, 4, 6, 7)[i])
+
+    # check proportion-too-fast all task trials:
+    rt = task_rt[0].append(task_rt[1]).append(task_rt[2]).append(task_rt[3])
+    prop_too_fast = len(rt[(rt < rt_FAST)]) / len(rt)
+    if prop_too_fast > threshold['rt_task_all']:
+        pct = 100 * prop_too_fast
+        warn += "%.0f%% trials with RT < %dms across all task blocks\n" % (
+            pct, rt_FASTms)
+
+    # check proportion-too-fast in each practice block:
+    for i, rt in enumerate(prac_rt):
+        prop_too_fast = len(rt[(rt < rt_FAST)]) / len(rt)
+        if prop_too_fast > threshold['rt_prac_all']:
+            pct = 100 * prop_too_fast
+            warn += "%.0f%% trials with RT < %dms in practice block #%d\n" % (
+                pct, rt_FASTms, (1, 2, 5)[i])
+
+    # check proportion-error in each practice block:
+    for i, prac_blk in enumerate(prac_ac):
+        if prac_blk.mean() < threshold['ac_prac_blk']:
+            pct = 100 * (1 - prac_blk.mean())
+            warn += "%.0f%% errors in practice block #%d\n" %(pct, (1, 2, 5)[i])
+
+    # check proportion-error in all practice trials:
+    ac = prac_ac[0].append(prac_ac[1]).append(prac_ac[2]).mean()
+    if ac < threshold['ac_prac_all']:
+        pct = 100 * (1 - ac.mean())
+        warn += "%.0f%% errors across all practice blocks\n" % pct
+
+    # check proportion-error in task blocks:
+    for i, ac in enumerate(task_ac):
+        if ac.mean() < threshold['ac_task_blk']:
+            pct = 100 * (1 - ac.mean())
+            warn += "%.0f%% errors in task block #%d\n" % (pct, (3, 4, 6, 7)[i])
+
+    # check proportion-error across all task trials:
+    ac = task_ac[0].append(task_ac[1]).append(task_ac[2]).append(task_ac[3])
+    if ac.mean() < threshold['ac_task_all']:
+        pct = 100 * (1 - ac.mean())
+        warn += "%.0f%% errors across all task trials\n" % pct
+
+    # ------------ Filter out bad trials: ------------
+    for i, block in enumerate(task_ac):
+        # retain trials with correct responses:
+        correct_trials = (block == correct)
+        task_rt[i] = task_rt[i][correct_trials]
+        #task_ac[i] = task_ac[i][correct_trials]
+    for i, block in enumerate(task_rt):
+        # retain trials where RT is not too fast or too slow:
+        rt_ok_trials = (block >= rt_FAST) & (block <= rt_SLOW)
+        task_rt[i] = task_rt[i][rt_ok_trials]
+        #task_ac[i] = task_ac[i][rt_ok_trials]
+
+    # ------------ Calculate summary stats of the filtered data: ----------
+    mean3, mean4, mean6, mean7 = [a.mean() for a in task_rt]
+    stdev36 = task_rt[0].append(task_rt[2]).std() # pooled std of blocks 3 & 6
+    stdev47 = task_rt[1].append(task_rt[3]).std() # pooled std of blocks 4 & 7
+    d36 = side * (mean6 - mean3) / stdev36  # side is +1 or -1
+    d47 = side * (mean7 - mean4) / stdev47
+    D_IAT = (d36 + d47) / 2
+
+    stats = D_IAT, side, mean3, mean4, mean6, mean7, stdev36, stdev47, warn.strip() or 'None'
+    labels = 'D_IAT', 'side', 'mean3', 'mean4', 'mean6', 'mean7', 'sd36', 'sd47', 'warnings'
+    if write_file:
+        df = pd.DataFrame([stats], columns=labels)
+        df.to_csv('Scored_' + csvfile, index=False, index_label=False, encoding='utf-8')
+
+    return warn.strip() or D_IAT
+
+def batchScoreIAT(path='.', write_file=False):
+    """Call scoreIAT() on all csv files in path
+    """
+    files = glob.glob(os.path.join(path, '*.csv'))
+    for f in files:
+        scoreIAT(f, write_file=write_file)
+
+if __name__ == '__main__':
+    for f in sys.argv[1:]:
+        print f, scoreIAT(f)