def residualize(X, ps):
numObs = X.shape[0]
resHat = np.zeros((numObs,1))
hOptimal = LLR.optimalBandwidthSelection(X, ps)
# for each observations, get prediction
for iObs in range(numObs):
xFit = LLR.polynomialFit(X,ps, ps[iObs], hOptimal)
resHat[iObs] = X[iObs] - xFit[0]
return resHat
def residualize(X, ps):
numObs = X.shape[0]
resHat = np.zeros((numObs, 1))
hOptimal = LLR.optimalBandwidthSelection(X, ps)
# for each observations, get prediction
for iObs in range(numObs):
xFit = LLR.polynomialFit(X, ps, ps[iObs], hOptimal)
resHat[iObs] = X[iObs] - xFit[0]
return resHat
def LLRKpLIVMTE(yTilda, ps, hOptimal, betaDif, x, u):
# test for size agree
if x.shape[1] != betaDif.shape[0]:
print 'the size of X and beta does not agree'
raise
elif x.shape[0] != 1:
print 'the x is not row vector'
raise
elif betaDif.shape[1] != 1:
print 'the beta is not column vector'
raise
result = LLR.polynomialFit(yTilda,ps,u,hOptimal)
mteHat = np.dot(x, betaDif) + result[2]
return mteHat
def LLRKpLIVMTE(yTilda, ps, hOptimal, betaDif, x, u):
# test for size agree
if x.shape[1] != betaDif.shape[0]:
print 'the size of X and beta does not agree'
raise
elif x.shape[0] != 1:
print 'the x is not row vector'
raise
elif betaDif.shape[1] != 1:
print 'the beta is not column vector'
raise
result = LLR.polynomialFit(yTilda, ps, u, hOptimal)
mteHat = np.dot(x, betaDif) + result[2]
return mteHat
