wq = pc.W(q)
# Set the paramter values into the element
dspr.setStiffness(yq['K'])
# Create space for fetching deterministic
# jacobian, and state vectors that go as input
uq = np.zeros((nddof))
lamq = np.zeros((nddof))
for k in range(nterms):
psiky = pc.evalOrthoNormalBasis(k, q)
uq[:] += U[k * nddof:(k + 1) * nddof] * psiky
lamq[:] += lam[k * nddof:(k + 1) * nddof] * psiky
# Evaluate function
dfdxq = dspr.getAdjointDeriv(uq, lamq)
# Project the determinic initial conditions onto the
# stochastic basis
psiziw = wq * pc.evalOrthoNormalBasis(i, q)
fitmp += dfdxq * psiziw
# Set into the force vector
dFdX[i] = fitmp
fmeanprime = dFdX[0]
fvarprime = np.sum(2 * E[1:] * dFdX[1:])
fstdprime = fvarprime / (2.0 * np.sqrt(fvar))
print('stochastic galerkin adjoint')
print("fmean : %15.14f" % fmean, "fmeanprime : %15.14f" % fmeanprime)
