Hi All,
I have run a binary response model using runmlwin, on multiple imputed data. I would like to report an R-squared value for the model fit but have not been able to find a reference on how to do this from runmlwin. Is there some option I could use to calculate this value or is there a different value that should be reported for model fit?
Many thanks,
Luis
R-squared for binary response model
-
GeorgeLeckie
- Site Admin
- Posts: 432
- Joined: Fri Apr 01, 2011 2:14 pm
Re: R-squared for binary response model
Dear Luis,
I'm afraid, there is no R-squared or equivalent statistic implemented in runmlwin.
Have a search for what people recommend for assessing fit in the multilevel logistic setting and then try calculating that statistic by hand.
For example, post-estimation you have access to the estimated model parameters and the empirical Bayes predictions of the random effects. Armed with these you can calculated predicted probabilities for each observation, which I'm guessing the various fit statistics which have been proposed are likely calculated.
Best wishes
George
I'm afraid, there is no R-squared or equivalent statistic implemented in runmlwin.
Have a search for what people recommend for assessing fit in the multilevel logistic setting and then try calculating that statistic by hand.
For example, post-estimation you have access to the estimated model parameters and the empirical Bayes predictions of the random effects. Armed with these you can calculated predicted probabilities for each observation, which I'm guessing the various fit statistics which have been proposed are likely calculated.
Best wishes
George