It means it's not able to compute the logarithm of the random effect
standard deviation, which is the parameter that's actually optimized
in the random effect term as opposed to the SD directly. Things to
try:
(1) Switch to -xtmelogit-.
(2) Be sure to run regular logistic regression first to get initial
values for the fixed effects coefficients. Stata does this already
automatically but you get information there. If your fixed effects
model is bad (e.g., it's at a separation point), then it's
unsurprising that it cannot determine a logsigma.
Beyond that not knowing the number of variables in the fixed effects,
it's tough to guess. You may have one or two highly unusual tracts
creating a crazy high initial estimate or vice versa essentially no
tract information left (due to use of fixed effects), or something
constant within a tract....
On Mon, Jan 28, 2013 at 12:31 PM, MA, XIAOGUANG MA <ma26@email.sc.edu> wrote:
> Hi,
>
> I am using xtlogit (re) to deal with my panel data (over 10,000 mothers in 150 census tracts). For several models, I got this error message and the xtlogit models rejected to run:
>
> "cannot compute lnsig2u initial value r(430);"
>
> What does this mean? Do I need to assign an initial value? How can I do that? How can I pass through this error and get the models run?
>
> Much appreciate.
>
> Max
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
--
JVVerkuilen, PhD
jvverkuilen@gmail.com
http://lesswrong.com/
"Everybody loves progress but nobody likes change." ---Fortune cookie, 1/13/13.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
