<>
Just try it out for the first question:
***
clear*
inp obs_rrt T unit
0 5 4
3 6 4
2 7 4
4 8 4
3 9 4
1 10 4
3 2 3
1 3 3
1 4 3
1 5 3
0 6 3
1 7 3
end
xtset unit T, monthly
//Does not converge
//xtpoisson obs_rrt, re
//Does converge, different model
xtpoisson obs_rrt T, re
***
The first version does not converge (thus commented out), the second one
does. -xtset- only serves to tell Stata that it has panel data in memory, so
you have to specify covariates separately...
HTH
Martin
_______________________
----- Original Message -----
From: "Mark Marshall" <[email protected]>
To: <[email protected]>
Sent: Saturday, April 11, 2009 11:37 PM
Subject: Corrected repost: Two general questions about xt commmands and re /
fe
Hi
I am using Stata 9.2 on windows.
I am fitting an xtpoisson model to unbalanced panel data (n counts of
death
over T months), with a grouping variable called unit that signifies the
ICU
where the data come from. E.g
obs_rrt T unit
0 5 4
3 6 4
2 7 4
4 8 4
3 9 4
1 10 4
3 2 3
1 3 3
1 4 3
1 5 3
0 6 3
1 7 3
Question 1.
When I specify
tis T
my understanding is that this does not explicitly model T in the model (I
want to model the death rate in relationship to time, amongst other
things),
and I would have to specify T in the command line to do this. Is this
correct?
Question 2.
I would like to specify T and some other exposure covariates as fixed
effects, and unit as a random effect in the model. If I specify all of
these
as independent variables in the command line, and then use the re option,
how can I tell which is modelled as fixed, and which is modelled as
random?
Thanks,
Mark Marshall
*
* 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/
