|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: Using univariate models to provide starting values for multivariate analysis
Katherine Boughey asks::
"Is it valid to use values from univariate analysis for starts in -xtmelogit-?"
It is absolutely valid to give your own starts, though -xtmelogit- is pretty darn good at picking starting values as it more or less does what you propose by fitting a set of simpler models, which are subsequently refined by simultaneous maximization.
If you are worried about local optima, you may want to force it to start from multiple locations.
JV
-----Original Message-----
From: [email protected]
To: [email protected]
Sent: 3/9/2009 8:46 AM
Subject: st: Using univariate models to provide starting values for multivariate analysis
Dear Statalisters,
I have a dataset that records bat abundance(converted for this analysis to
presence/absence) along 30km transects. Each transect is surveyed twice a
year giving two levels of clustering: surveys nested in transects and
transects nested within year.
I hope to model the relationship between bat presence and a variety of
explanatory predictors using -xtmelogit-. I have developed a set of
candidate models each containing a different combination of explanatory
predictors.
In a previous analysis I fitted univariate models for each explanatory
predictor. Is it valid to use these estimates as a starting value for that
predictor in each candidate model in which that predictor appears?
Thanks,
Katherine Boughey
--
Centre for Ecology, Evolution and Conservation
School of Environmental Sciences
University of East Anglia
Norwich
NR4 7TJ
*
* 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/
*
* 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/