[Gretl-users] Re: A question on Kalman Filter/MLE
Am 20.04.2020 um 21:16 schrieb Alecos Papadopoulos:
Given starting values for the parameters, and all the rest, the
Kalman
filter runs, and it gives the first log-likelihood value series. Then
the MLE takes over, but it immediately finds that
/"matrix is not positive definite"./
*The question is: which matrix is the message referring to?*
loglikelihood = -257.320305306
Parameters: -0.98707 0.75974 -2.5877 -0.46878
...
Gradients: 0.00000 0.00000 0.00000 0.00000 0.00000
...
Tolerance = 1.81899e-012
Matrix is not positive definite
Error executing script: halting
Not being the author of the Kalman code and without further
investigation I can only speculate (and I agree that a more informative
error message would be good), but it would help if we could see more of
your script that you're using.
Having said that, my speculation would be that it refers to the
estimation of VCV. To diagnose this perhaps experiment with --hessian or
--auxiliary in the mle step.
cheers
sven