www.cmm.bristol.ac.uk/forum

Dear all

I am planning to model my data (3-level , with ordered response var.) in runmlwin. I only faced to missing for variabels in level 1. Now, I have few quastiones; first, I used the stata to impute missing values(multiple imputation), is it true or is there any problem with that? second, how can I check the proportionality as the assumption of ordinal logit modes?

I am looking for your advises,
Regards, Ali

Dear Ali,

Yes in general one can create multiple imputed datasets by some appropriate procedure and then use runmlwin to for the model of interest to each of these datasets. One can use the -mi- site of commands to combine the model results using Rubin's rules. One thing to bear in mind is that the imputation model should be at least as complex the as the model of interest. So if the model of interest is multilevel so should be the imputation model.

You can test the proportional odds model by relaxing it via allowing the coefficients to vary across the log odds contrasts. See the MLwiN user manual for further details.

Best wishes

George

Dear George
Manay many thanks.
Regards