def fcit_wrapper(data_matrix, i, j, ks, **kwargs):
x = data_matrix[:, i].reshape(-1, 1)
y = data_matrix[:, j].reshape(-1, 1)
if len(ks) == 0:
res = fcit.test(x, y)
else:
res = fcit.test(x, y, data_matrix[:, list(ks)])
return 1.0 if np.isnan(res) else res
def testFCIT(ps, x, y, z=[]):
import numpy as np
from fcit import fcit
X = [ps.ds[x[0]]]
Y = [ps.ds[y[0]]]
Z = []
for var in z:
zdat = ps.ds[var]
Z.append(zdat)
Xa = np.array(X).transpose()
#print('xshape = ', Xa.shape)
Ya = np.array(Y).transpose()
if Z:
Za = np.array(Z).transpose()
#print('zshape = ', Za.shape)
pval = fcit.test(Xa, Ya, Za, num_perm=10, prop_test=.40)
else:
pval = fcit.test(Xa, Ya, num_perm=10, prop_test=.40)
return pval
def nonparametric_fcit_test(X, Y, Z, data):
"""
X and Y are names of variables.
Z is a list of names of variables.
data is a pandas data frame.
Return a float corresponding to the p-value computed from FCIT.
"""
x_np = np.transpose(np.asmatrix(data[X].values))
y_np = np.transpose(np.asmatrix(data[Y].values))
z_np = np.asmatrix(data[Z].values)
p = fcit.test(x_np, y_np, z_np)
return p
def nonparametric_fcit_test(X, Y, Z, data):
"""
X and Y are names of variables.
Z is a list of names of variables.
data is a pandas data frame.
Return a float corresponding to the p-value computed from FCIT.
"""
# implement code here
X_var = np.asmatrix(data[X].to_numpy()).transpose()
Y_var = np.asmatrix(data[Y].to_numpy()).transpose()
Z_var = np.asmatrix(data[Z].to_numpy())
print(X_var.shape, Y_var.shape, Z_var.shape)
pval = fcit.test(X_var, Y_var, Z_var)
return pval
def statistic_DFA(X,T=None, p0 = 5e-2, p1 = 0.1, n_perm = 15, n_itr=5,n_samp=250, verbos=False):
if T is None:
T = np.linspace(0,1,X.shape[0])
if len(T.shape) == 1:
T = T.reshape(-1,1)
nd,fd,di = list(range(X.shape[1])),[],[]
if verbos:
print("Searching all drifting features")
rel = RIT(T,X,p0, n_itr,n_samp)
if verbos:
print("Searching drift inducing features")
for i_num,i in enumerate(rel):
if verbos:
print(i_num,"/",len(rel))
nd.remove(i)
R_i = np.array(np.hstack((range(0,i),range(i+1,X.shape[1]))),dtype=int)
if fcit.test(T,X[:,i].reshape(-1,1),X[:,R_i],n_perm=n_perm) < p1:
di.append(i)
else:
fd.append(i)
return {"non-drifting":nd,"faithfully-drifting":fd,"drift-inducing":di}