The over all variation explained by the model

[davedan8]

Hi guys!
Last time i posted a question regarding the Goodness-of-fit of the model when IGLS is used but didn't get any reply on it. While i was referring to Snijders and Bosker, i read the R-square equivalent of the multilevel analysis in binary response model. The idea put there is: say if you denote the variance of the fixed part as Sigma-square F , that of the random part for the higher level as Sigma-square U and for the lower level as Sgma-sqaure e (in this case 3.29), then Obviously the sum total of these parameters is the over all variation in the model. The book went on discussing how to calculate the R-square as:
R-square= Sigma-square F /(Sigma-square F + Sigma-square U + Sigma-square e) where Sigma-square e is still 3.29
The theoretical foundation sounds quite good but i couldn't get how to implement this in MLwin. Does any one know how it is implemented in MLwin? or any other equivalent mechanism that enables to know the over all Goodness-of-fit of the model in 2 level logit model when the IGLS estimation procedure is used?
Thanks a lot!!
Dave

[bosawas]

I tend to use MCMC estimation and DIC, but I'd be curious to learn about ways to assess model fit in MLwiN for a logistic regression.

[davedan8]

Tnx Bosawas for the reply!
oh yeah! it gives a DIC result when MCMC is used but nothing when IGLS is used. As a coincidence, am not using MCMC, that's why i asked. but am also eager to know more on Model-fit-Goodness in MLwin.
bye!