To see the differences of variances explained by two models. Because most researchers use the additional software like MLwin etc.
Is this the same as a GLM procedure? that gives significance levels
I'm not sure I'm understanding you correctly, but it sounds like your confusing tests of relative model fit (e.g., log-likelihood ratio test), which would be used in comparing null and "final" models ("final" as in the final variance model or as in the final means model with all your predictors?), with a measure of relative effect size (i.e., proportion of variance accounted at a specific level of analysis). If you're looking for a test of model fit, I don't think any stats package provides a correct hypothesis test of model fit of the sort you likely need (i.e., log-likelihood ratio test for unbalanced and missing data), though, they should all provide you with the information you need to calculate the correct statistic by hand (i.e., -2LL, # of parameters). If you're looking to describe (not test) the proportion of variance explained by your predictors (predictors explain variance; random effects partition variance, not explain it, meaning your null model can't explain variance), then this also must be calculated by hand, usually as Pseudo-R2.
In either case, I've uploaded an Excel document that cranks out the math for you in order to compute log-likelihood ratio tests, Pseudo-R2, and random effects confidence intervals. All you need to enter into it is -2LL, # of parameters, and the estimates for each of your variance components, all of which should be provided in the output of any self-respecting stat package. It's uploaded as a "dataset".
Hope this helps
Hi,
For your question, kindly see the attachment pdf.
thank you.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS