Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: Integer log likelihoods
From
Jorge Eduardo Pérez Pérez <[email protected]>
To
<[email protected]>
Subject
st: Integer log likelihoods
Date
Wed, 13 Apr 2011 12:39:42 -0400
Dear Statalist,
I am trying to estimate an unobserved components time series model
using the Kalman Filter. I am using Stata MP 11.2. My model is
non-stationary and a little ill conditioned, so I expected
optimization problems. However, I am getting rather strange results:
log likelihoods become integer after a few iterations ! I am using the
-diff- option in the optimization. If I don't use it, the problem does
not arise.
Here is the code and some of the output. I was not able to reproduce
the problem with a Stata dataset.
constraint drop _all
constraint 1 [mu]l.mu=1
constraint 2 [lc1]l.c1=1
constraint 3 [dlrgdp]mu=1
constraint 4 [dlrgdp]c1=1
foreach x in "Argentina" {
sspace (mu l.mu e.mu, state noconstant) (c1 l.c1 l.lc1 e.c1, state
noconstant) (lc1 l.c1, state noconstant) (dlrgdp mu c1 e.dlrgdp) if
country=="`x'", covstate(identity) constraints(1/4) diff
}
Iteration 0: log likelihood = 106.07584
Iteration 1: log likelihood = 106.37785
Iteration 2: log likelihood = 106.37786 (backed up)
Iteration 3: log likelihood = 106.37786 (backed up)
Iteration 4: log likelihood = 106.37787 (backed up)
(switching technique to nr)
Iteration 5: log likelihood = 106.37787 (not concave)
Iteration 6: log likelihood = 106.84663 (not concave)
Iteration 7: log likelihood = 108.23049 (not concave)
Iteration 8: log likelihood = 144 (not concave)
Iteration 9: log likelihood = 416 (not concave)
Iteration 10: log likelihood = 416 (not concave)
Iteration 11: log likelihood = 416 (not concave)
Iteration 12: log likelihood = 416 (not concave)
Iteration 13: log likelihood = 244 (not concave)
I think I should submit this to Tech Support, but I wanted to see if
somebody had suggestions on this matter.
_______________________
Jorge Eduardo Pérez Pérez
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/