def test_hypothesis_3():
# new multiple-regression
img = nib.load(pathtoclassdata + "ds114_sub009_t2r1.nii")
data = img.get_data()[..., 4:]
# Read in the convolutions.
convolved = np.loadtxt(pathtoclassdata + "ds114_sub009_t2r1_conv.txt")[4:]
# Create design matrix.
X = np.ones((convolved.shape[0], 2))
X[:, 1] = convolved
beta, t, df, p = t_stat(data, convolved, [0, 1])
beta2, t2, df2, p2 = t_stat_mult_regression_single(data, X)
beta3, t3, df3, p3 = t_stat_mult_regression(data, X)
assert_array_equal(t, t2)
assert_array_equal(t, np.atleast_2d(t3[1, :]))
def test_hypothesis_3():
# new multiple-regression
img = nib.load(pathtoclassdata + "ds114_sub009_t2r1.nii")
data = img.get_data()[..., 4:]
# Read in the convolutions.
convolved = np.loadtxt(pathtoclassdata + "ds114_sub009_t2r1_conv.txt")[4:]
# Create design matrix.
X=np.ones((convolved.shape[0],2))
X[:,1]=convolved
beta,t,df,p = t_stat(data, convolved,[0,1])
beta2, t2,df2,p2 = t_stat_mult_regression_single(data, X)
beta3, t3,df3,p3 = t_stat_mult_regression(data, X)
assert_array_equal(t,t2)
assert_array_equal(t,np.atleast_2d(t3[1,:]))
X_cond = np.ones((n_vols, 17))
X_cond[:, 1] = hrf_matrix_1[:, j] # 1 more
X_cond[:, 2] = hrf_matrix_2[:, j] # 1 more
X_cond[:, 3] = hrf_matrix_3[:, j] # 1 more
X_cond[:, 4] = np.linspace(-1, 1, num=X.shape[0]) #drift # one
X_cond[:, 5:11] = fourier_creation(X.shape[0], 3)[:, 1:] # six more
X_cond[:, 11:] = pca_addition
#START CREATING MODELS
###################
# MODEL 1 #
###################
# hrf (simple)
beta1, t, df1, p = t_stat_mult_regression(data_slice, X[:, 0:2])
MRSS1, fitted, residuals = glm_diagnostics(beta1, X[:, 0:2],
data_slice)
model1_slice = np.zeros(len(MRSS1))
rank1 = npl.matrix_rank(X[:, 0:2])
count = 0
for value in MRSS1:
model1_slice[count] = adjR2(value, np.array(data_slice[count, :]),
df1, rank1)
count += 1
#Run per slice in order to correct for time
for j in range(data.shape[2]):
data_slice = data[:, :, j, :]
#Create design matrix
X = np.ones((n_vols, 9))
X[:, 1] = convolve[:, j]
X[:, 2] = np.linspace(-1, 1, num = X.shape[0]) #drift
X[:, 3:] = fourier_creation(n_vols, 3)[:, 1:]
beta, t, df, p = t_stat_mult_regression(data_slice, X)
#take only first coefficient
t = t[1, :]
p = p[1, :]
MRSS, fitted, residuals = glm_diagnostics(beta, X, data_slice)
beta = beta[:, 1]
#insert into the proper slice
beta_final[:, :, j] = beta.reshape(data_slice.shape[:-1])
t_final[:, :, j] = t.reshape(data_slice.shape[:-1])
p_final[:, :, j] = p.reshape(data_slice.shape[:-1])
residual_final[:, :, j, :] = residuals.reshape(data_slice.shape)
X_cond[:, 1] = hrf_matrix_1[:, j] # 1 more
X_cond[:, 2] = hrf_matrix_2[:, j] # 1 more
X_cond[:, 3] = hrf_matrix_3[:, j] # 1 more
X_cond[:, 4] = np.linspace(-1, 1, num = X.shape[0]) #drift # one
X_cond[:, 5:11] = fourier_creation(X.shape[0], 3)[:, 1:] # six more
X_cond[:, 11:] = pca_addition
#START CREATING MODELS
###################
# MODEL 1 #
###################
# hrf (simple)
beta1, t,df1, p = t_stat_mult_regression(data_slice, X[:, 0:2])
MRSS1, fitted, residuals = glm_diagnostics(beta1, X[:, 0:2], data_slice)
model1_slice = np.zeros(len(MRSS1))
rank1 = npl.matrix_rank(X[:, 0:2])
count = 0
for value in MRSS1:
model1_slice[count] = adjR2(value, np.array(data_slice[count, :]), df1, rank1)
count += 1
adjr2_1 = adjr2_1 + model1_slice.tolist()
n_vols = data.shape[-1]
convolve = np.loadtxt(hrf_data + i + "_hrf.txt")
residual_final = np.zeros((data.shape))
t_final = np.zeros((data.shape[:-1]))
for j in range(data.shape[2]):
data_slice = data[:, :, j, :]
X = np.ones((n_vols, 6))
X[:, 1] = convolve[:, j]
X[:, 2] = np.linspace(-1, 1, num=X.shape[0]) #drift
X[:, 3:] = fourier_creation(X.shape[0], 3)[:, 1:]
beta, t, df, p = t_stat_mult_regression(data_slice, X)
t = t[1, :]
MRSS, fitted, residuals = glm_diagnostics(beta, X, data_slice)
t_final[:, :, j] = t.reshape(data_slice.shape[:-1])
residual_final[:, :, j, :] = residuals.reshape(data_slice.shape)
np.save("../data/glm/t_stat/" + i + "_tstat.npy", t_final)
np.save("../data/glm/residual/" + i + "_residual.npy", residual_final)
sys.stdout.write("-")
sys.stdout.flush()
sys.stdout.write("\n")
t_final1 = np.zeros((data_smooth.shape[:-1]))
t_final2 = np.zeros((data_smooth.shape[:-1]))
t_final3 = np.zeros((data_smooth.shape[:-1]))
#Run per slice in order to correct for time
for j in range(data_smooth.shape[2]):
data_smooth_slice = data_smooth[:,:,j,:]
data_rough_slice = data_rough[:,:,j,:]
#Create design matrix
X = np.ones((n_vols,7))
X[:,1] = convolve[:,j]
X[:,2]=np.linspace(-1,1,num=X.shape[0]) #drift
X[:,3:]=fourier_creation(X.shape[0],2)[:,1:]
beta1,t1,df1,p1 = t_stat_mult_regression(data_rough_slice, X)
beta2, t2, df2, p2 = t_stat(data_smooth_slice,convolve[:,j], c=[0,1] )
beta3, t3, df3, p3 = t_stat(data_rough_slice,convolve[:,j], c=[0,1] )
t1 = t1[1,:]
t2 = t2.T
t3= t3.T
t_final1[:,:,j] = t1.reshape(data_rough_slice.shape[:-1])
t_final2[:,:,j] = t2.reshape(data_smooth_slice.shape[:-1])
t_final3[:,:,j] = t3.reshape(data_rough_slice.shape[:-1])