Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.
As Nick has pointed out, you have not provided anything close to
sufficient information for respondents to provide much in the way of
To provide the context: -xtpmg- is a user-written command that relies
upon the Stata maximum likelihood procedure -ml-. Infeasible
starting values for maximum likelihood procedures are difficult to
deal with under the best of circumstances. Hence, you will need to
consider in detail why they might have arisen or what can be
done. For example, are there missing or extreme values in your data
or is the data ill-conditioned (collinearity)? Does the problem
arise with a simpler model?
More generally, you cannot treat a procedure like this as a black
box. You may have to look at the details of the xtpmg.ado file and
then modify it to create a version that works for you. Many -ml-
procedures allow the from() option to specify starting values rather
than allowing the program to generate them. It is not clear whether
this is possible with -xtpmg- since the help file seems to be
contradictory, but you can change the manner in which the procedure
creates its starting values.
Finally, I note that there is a Stata Journal article linked to the
procedure as well as a variety of estimation options which may not
suffer from the same problem. There are also other user-written
procedures which estimate mean group models - e.g. -xtmg- written by
Marcus Eberhardt and which is not based on use of -ml-.