Dear statalist-users,
BACKGROUND:
I am analyzing a longitudinal panel of persons in a fixed effect framework (FE) that needs to be pweighted. I am fitting five series of models (e.g. differing time-points, because of availability of control variables) with four different dependent variables each. Three of the dependent variables are continious (for those I use <xtreg>) and one is binary.
For the binary dependent variable I am using <clogit> because <xtlogit> does not allow pweights, while clogit does. For my final results table, I would like to give for each model the number of observations (person*year) and the number of groups (persons), like given in the xt-estimation results. I want to do this, because they differ depending on the control variables included and item-nonresponse with those controls.
MY PROBLEM:
As far as I was able to find out, clogit only returns
e(N) = no. of obs., in my case person*year
e(N_drop) = no. of observations that were dropped, because of all pos./neg. outcome
e(N_group_drop) = the no. of groups (in my case persons) that were dropped because of all pos./neg. outcome
I do not think that I can infer from the number of people in my panel minus e(N_group_drop) how many people were used in the specific estimation (because of item-nonresponse and differentially specificated models). That is why I would like to know, if anyone can think of a reliable, and maybe even convenient, way of finding "N_group"?
///Please apologise, if this is straightforward! I only do my second analysis with stata...
Best regards
Vanita Matta
BACKGROUND:
I am analyzing a longitudinal panel of persons in a fixed effect framework (FE) that needs to be pweighted. I am fitting five series of models (e.g. differing time-points, because of availability of control variables) with four different dependent variables each. Three of the dependent variables are continious (for those I use <xtreg>) and one is binary.
For the binary dependent variable I am using <clogit> because <xtlogit> does not allow pweights, while clogit does. For my final results table, I would like to give for each model the number of observations (person*year) and the number of groups (persons), like given in the xt-estimation results. I want to do this, because they differ depending on the control variables included and item-nonresponse with those controls.
MY PROBLEM:
As far as I was able to find out, clogit only returns
e(N) = no. of obs., in my case person*year
e(N_drop) = no. of observations that were dropped, because of all pos./neg. outcome
e(N_group_drop) = the no. of groups (in my case persons) that were dropped because of all pos./neg. outcome
I do not think that I can infer from the number of people in my panel minus e(N_group_drop) how many people were used in the specific estimation (because of item-nonresponse and differentially specificated models). That is why I would like to know, if anyone can think of a reliable, and maybe even convenient, way of finding "N_group"?
///Please apologise, if this is straightforward! I only do my second analysis with stata...
Best regards
Vanita Matta
