Hi,
I would like to assess the effect of GP-gender on receiving care (for patients (binary outcome, received=1, did not received=0)
adjusted for characteristics of patients (agegroup, gender), phyisican (age, gender) and practice (location, size).
As the outcome is binary, I am thinking on logistic model.
The patients are clustered within practices therefore adjustment of the model for the clustering effect is necessary.
The number of physicians are relatively high (N=4575).
I used conditional logistic regression grouped by GP-ID (clogit DEP_var INDEP_var, group(GP ID) robust).
The problem is that after running the model I have got the following error: "var omitted because of no within-group variance".
So most of the characteristics are dropped because characteristics don’t vary within GPs (e.g. characteristics of practice or GP).
The problem is the same with ’xtlogit’. The data were cross-sectional so the changes over time were not monitored (no panel data).
What should I do? Can I use random-effect here?
Best, Monika
I would like to assess the effect of GP-gender on receiving care (for patients (binary outcome, received=1, did not received=0)
adjusted for characteristics of patients (agegroup, gender), phyisican (age, gender) and practice (location, size).
As the outcome is binary, I am thinking on logistic model.
The patients are clustered within practices therefore adjustment of the model for the clustering effect is necessary.
The number of physicians are relatively high (N=4575).
I used conditional logistic regression grouped by GP-ID (clogit DEP_var INDEP_var, group(GP ID) robust).
The problem is that after running the model I have got the following error: "var omitted because of no within-group variance".
So most of the characteristics are dropped because characteristics don’t vary within GPs (e.g. characteristics of practice or GP).
The problem is the same with ’xtlogit’. The data were cross-sectional so the changes over time were not monitored (no panel data).
What should I do? Can I use random-effect here?
Best, Monika
Comment