www.cmm.bristol.ac.uk/forum

Dear all,

I am modelling a single level unordered multinomial logistic model in r2mlwin. Because my model is single level, I have to estimate it in first order MQL. In MQL, however, the likelihood value is not shown. To compare my results I would like to obtain a likelihood value (on the website I found that is possible to display the likelihood value for IGLS estimation).

Is it possible to obtain such a measure (or an approximation) in MQL? And if so, what code do i need to enter?

This is an example of the simple code I am currently using:

model_a = logit(dep, denom) ~ 1 + var1 + var2
model_a<- runMLwiN(model_a, D = "Unordered Multinomial", data =mydata)

Kind regards,

Pim Verbunt

Unfortunately as discrete response models estimated in MLwiN using (R)IGlS use quasi-likelihood it is not possible to obtain an accurate log-likelihood. For an explanation in the context of the -runmlwin- command see https://www.cmm.bristol.ac.uk/forum/viewtopic.php?t=279.