def calcMCM(self):
''' compute model covariance matrix '''
J = gmat2numpy( self.f.jacobian() ) # (linear) jacobian matrix
D = np.diag( 1 / self.error )
DJ = D.dot( J )
JTJ = DJ.T.dot( DJ )
MCM = inv( JTJ ) # model covariance matrix
varVG = np.sqrt( np.diag( MCM ) ) # standard deviations from main diagonal
di = ( 1. / varVG ) # variances as column vector
MCMs = di.reshape(len(di),1) * MCM * di # scaled model covariance (=correlation) matrix
return varVG, MCMs
def calcMCM(self):
"""Compute linear model covariance matrix."""
J = gmat2numpy(self.fop.jacobian()) # (linear) jacobian matrix
D = np.diag(1 / self.error)
DJ = D.dot(J)
JTJ = DJ.T.dot(DJ)
MCM = np.linalg.inv(JTJ) # model covariance matrix
var = np.sqrt(np.diag(MCM)) # standard deviations from main diagonal
di = (1. / var) # variances as column vector
# scaled model covariance (=correlation) matrix
MCMs = di.reshape(len(di), 1) * MCM * di
return var, MCMs