www.cmm.bristol.ac.uk/forum

Hi,

I'm fitting a 2 level random intercept logistic model with a cross level interaction term. I've recently started using runmlwin instead of MLwin and am having trouble replicating the graphical depiction of this as per module 7 (pg 52 on). When attempting the command predict predprob, fixed only
stata reports it can't find my explanatory variables.

I'm wondering if this is because I am using the i.var command for each of my explantory variables (they are all categorical except for my single continuous level 2 explanatory variable) and then how I best get around this (without creating dummy variables for all of my explanatory variables, which have done for my cross level interaction terms).

for eg educat =no school cert, school cert, diploma, university
and is reported by stata post MLwin as _1b_educat _2_educat _3_educat etc
and the error comes back as
variable _1b_educat not found

Thanks so much
Danielle

If you are interested in a comparison then http://www.bristol.ac.uk/cmm/software/r ... /examples/ replicates the models in LEMMA module 7 using -runmlwin-.

The dummy variables used for estimation are created as temporary variables in -runmlwin- using the http://www.stata.com/help.cgi?fvexpand and http://www.stata.com/help.cgi?fvrevar Stata commands. As the data created by these doesn't exist after running the command later commands will not be able to find them. I suspect that if you didn't want to manually create the dummies that you could use these commands to recreate what was used during estimation.