def compute_statistic(self, alphahat, R, RA, N, Nref, memoize=False):
# TODO: should we regularize RA?
print("regularizing R...")
Rreg = R.add_ridge(self.params.Lambda, renormalize=True)
if not memoize or not hasattr(self, "bias"):
print("done.computing bias...")
self.bias = BlockDiag.solve(Rreg, RA).trace() / N
print("bias =", self.bias)
betahat = BlockDiag.solve(Rreg, alphahat)
return betahat.dot(RA.dot(betahat)) - self.bias
def compute_statistic(self, alphahat, R, RA, N, Nref, memoize=False):
#TODO: should we regularize RA?
print('regularizing R...')
Rreg = R.add_ridge(self.params.Lambda, renormalize=True)
if not memoize or not hasattr(self, 'bias'):
print('done.computing bias...')
self.bias = BlockDiag.solve(Rreg, RA).trace() / N
print('bias =', self.bias)
betahat = BlockDiag.solve(Rreg, alphahat)
return betahat.dot(RA.dot(betahat)) - self.bias
def compute_statistic(self, alphahat, R, RA, N, Nref, memoize=False):
try:
if not memoize or not hasattr(self, "bias"):
print("computing bias")
self.bias = BlockDiag.solve(R, RA).trace() / N
print("bias =", self.bias)
betahat = BlockDiag.solve(R, alphahat)
return betahat.dot(RA.dot(betahat)) - self.bias
except np.linalg.linalg.LinAlgError:
print("R was singular. Its shape was", R.shape(), "and Nref=", Nref)
return 0
def compute_statistic(self, alphahat, R, RA, N, Nref, memoize=False):
try:
if not memoize or not hasattr(self, 'bias'):
print('computing bias')
self.bias = BlockDiag.solve(R, RA).trace() / N
print('bias =', self.bias)
betahat = BlockDiag.solve(R, alphahat)
return betahat.dot(RA.dot(betahat)) - self.bias
except np.linalg.linalg.LinAlgError:
print('R was singular. Its shape was', R.shape(), 'and Nref=',
Nref)
return 0
