 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: heteroskedastic multivariate normal regression with ml, convergence problem
|
From |
"Verkuilen, Jay" <[email protected]> |
|
To |
<[email protected]> |
|
Subject |
RE: st: heteroskedastic multivariate normal regression with ml, convergence problem |
|
Date |
Mon, 23 Feb 2009 14:47:23 -0500 |
Carlo Fezzi wrote:
>>In fact I tried increasing the number of observations to 10000 and the
algorithm converges easily after roughly 40 iterations... I guess more
experimentation is needed to find general rules..<<
Sounds like you have an empirical unidentification problem, at least in
this sample, which doesn't surprise me at all. Identifying higher order
moments well is not easy.
I'd say using the homoscedastic model as a good start for the location
model with the log-MSE as the start for the variance model will probably
help a lot but you should still expect to need large sample sizes.
JV
*
* 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/