I have a piece of -ml- code that produces something like
initial: log likelihood = -48417.771
improve: log likelihood = -3801.8795
rescale: log likelihood = -3533.3397
rescale eq: log likelihood = -1917.3358
Iteration 0: log likelihood = -4123.2086 (not concave)
Iteration 1: log likelihood = -4054.4958 (not concave)
or even something like
initial: log likelihood = -1107.3744
rescale: log likelihood = -1107.3744
rescale eq: log likelihood = -810.07783
initial values not feasible
r(1400);
I suspect the reason is that within my ado file, I specify some -ml
init- statements (or equivalent -init()- options of -ml, maximize-) --
but I still was expecting Stata to make the transition and use the
most recent and the best values found. In fact, in the last case, the
vector of -e(b)- was all zeroes, so both my default starting values
and the found values were ignored!
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
* 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/
