Merge 6608126 into f79fc9f

code/utils/linear_modeling.py

@@ -0,0 +1,137 @@
+# Python 3 compatibility
+from __future__ import absolute_import, division, print_function 
+import numpy as np
+import matplotlib.pyplot as plt
+import numpy.linalg as npl
+from scipy.stats import t as t_dist
+
+import scene_slicer
+import plot
+
+def get_design_matrix(data):                                                   
+    """
+    Returns
+    -------
+    Design matrix with 5 columns, including the 3 columns of interest,
+    the linear drift column, and the column of ones
+    """
+    n_trs = data.shape[-1]
+    X = np.ones((n_trs,5))   
+    day, inter, pos = data.scene_slicer.get_day_night()
+    X[:,0] = day
+    X[:,1] = inter
+    X[:,2] = pos
+    X[:,3] = np.linspace(-1, 1, n_trs)
+    return X
+def plot_design_matrix(X):
+    """
+    Returns
+    -------
+    Nothing
+    """
+    plt.imshow(X,aspect = 0.1, cmap='gray',interpolation='nearest')
+def get_betas(X,data):
+    Y = np.reshape(data,(data.shape[-1],-1))
+    B = npl.pinv(X).dot(Y)
+    return B
+def get_betas_4d(B,data):
+    """
+    Returns
+    -------
+    4D array, beta for each voxel
+    """
+    b_vols = np.reshape(B, data.shape[:-1] + (-1))  
+    return b_vols
+def plot_betas(b_vols,col):
+    """
+    Returns
+    -------
+    Nothing
+    """
+    plot.plot_central_slice(b_vols[...,col])
+def t_stat(y, X, c):
+    """ betas, t statistic and significance test given data, 
+    design matix, contrast
+    This is OLS estimation; we assume the errors to have independent
+    and identical normal distributions around zero for each $i$ in
+    $\e_i$ (i.i.d).
+    """
+    # Make sure y, X, c are all arrays
+    y = np.asarray(y)
+    X = np.asarray(X)
+    c = np.atleast_2d(c).T  # As column vector
+    # Calculate the parameters - b hat
+    beta = npl.pinv(X).dot(y)
+    # The fitted values - y hat
+    fitted = X.dot(beta)
+    # Residual error
+    errors = y - fitted
+    # Residual sum of squares
+    RSS = (errors**2).sum(axis=0)
+    # Degrees of freedom is the number of observations n minus the number
+    # of independent regressors we have used.  If all the regressor
+    # columns in X are independent then the (matrix rank of X) == p
+    # (where p the number of columns in X). If there is one column that
+    # can be expressed as a linear sum of the other columns then
+    # (matrix rank of X) will be p - 1 - and so on.
+    df = X.shape[0] - npl.matrix_rank(X)
+    # Mean residual sum of squares
+    MRSS = RSS / df
+    # calculate bottom half of t statistic
+    SE = np.sqrt(MRSS * c.T.dot(npl.pinv(X.T.dot(X)).dot(c)))
+    t = c.T.dot(beta) / SE
+    # Get p value for t value using cumulative density dunction
+    # (CDF) of t distribution
+    ltp = t_dist.cdf(t, df) # lower tail p
+    p = 1 - ltp # upper tail p
+    return beta, t, df, p
+def get_ts(Y,X,c,data):
+    """
+    Parameters
+    ----------
+    Y: TxB array, last axis of data
+    X: design matrix
+    c: contrast array
+
+    Returns
+    -------
+    t-statistic for each voxel
+    """
+    n_voxels = np.prod(data.shape[:-1])
+    t = np.zeros(n_voxels)
+    for num in range(n_voxels):
+	t[num] = t_stat(Y[:,num],X,c)[1]
+    return t
+
+def get_top_20(t):
+    """
+    Parameters
+    ----------
+    t: 1D array of t-statistics for each voxel
+
+    Returns
+    -------
+    1D array of position of voxels in top 20% of t-statistics (all are 
+    positive)
+    """
+    a = np.int32(round(len(t)*.2))
+    top_20_voxels = np.argpartition(t, -a)[-a:]
+    return top_20_voxels
+
+def get_index_4d(data,top_20_voxels):
+    """
+    Returns
+    -------
+    Indices in terms of 4D array of each voxel in top 20% of t-statistics
+    """
+    shape = data[...,-1].shape
+    axes = np.unravel_index(top_20_voxels,shape)
+    return zip(axes[0],axes[1],axes[2])
+def plot_single_voxel(data,top_20_voxels):
+    """
+    Returns
+    -------
+    Nothing
+    """
+    plt.plot(data[get_index_4d(data,top_20_voxels)[0]])
+