I am working on fitting a mixed model with logit link and survey adjustment. The code for a model with a categorical and continuous variable is
svy linearized: melogit dependent c.continuous i.categorical || randomeffect:, covariance(unstructured)
I get estimates for the coefficients but no errors or p-values. The model runs without problems if I either leave out the svy command or the random effect. It will also run with just the categorical variable. Any suggestions on how to make this work?
My other problem is how to compare models. Likelihood ratio test are not possible with survey estimations ("lrtest is not appropriate with survey estimation results"*/) and estat ic to obtain AIC is also not possible with this type of model. (/*"invalid subcommand ic"*/).
Thanks for any comments.
svy linearized: melogit dependent c.continuous i.categorical || randomeffect:, covariance(unstructured)
I get estimates for the coefficients but no errors or p-values. The model runs without problems if I either leave out the svy command or the random effect. It will also run with just the categorical variable. Any suggestions on how to make this work?
My other problem is how to compare models. Likelihood ratio test are not possible with survey estimations ("lrtest is not appropriate with survey estimation results"*/) and estat ic to obtain AIC is also not possible with this type of model. (/*"invalid subcommand ic"*/).
Thanks for any comments.
Comment