In general, remote diagnosis is difficult for this kind of problem. A
visceral suspicion is that you may be trying to fit a model much more
complicated than the data are willing to bear. From your account, you
are fitting in a high-dimensional parameter space, even if you have lots
of observations.
To steal one of Maarten's lines: start with something much simpler and
see whether it too is problematic. If not, build up to this model,
paying attention to variables that don't seem to help.
Nick
[email protected]
Jennifer Schmitt
I am running the following model
xi. xtmelogit knew_SNP i.villagecode i.area i.tribegroup snp_distance
cpa_distance age sex etc (I have another 25+ variables, mostly
dichotomous) || subvillage_code: || household_id:
I then get an "initial values not feasible r(1400)" error. I tried to
work around the problem by telling STATA to start with initial values
from a simple logit command (using the "-from()-" command), but it still
reports the error. I also considered extreme values, but those are not
an issue. Any thoughts or suggestions would be greatly appreciated!
*
* 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/
