Search found 156 matches

by billb
Thu Jan 23, 2020 3:07 pm
Forum: MLwiN user forum
Topic: Is level 2 variance significantly different from 0?
Replies: 1
Views: 3082

Re: Is level 2 variance significantly different from 0?

That depends on which estimation method you are using but generally involves fitting the model with and without the random intercept and looking at some indicator of model fit - a LR test for ML methods or comparing DIC for MCMC. Suggest you look in the user guides for examples.
Bill.
by billb
Thu Jan 23, 2020 3:05 pm
Forum: runmlwin user forum
Topic: cross-classified MLM - MCM - random and fixed effects
Replies: 1
Views: 3317

Re: cross-classified MLM - MCM - random and fixed effects

Hi Johann, Generally when we talk about cross-classified models we are assuming that each of these levels is a random source of variation and is treated as a set of RANDOM effects. It doesn't make a lot of sense to include fixed effects as well as random effects for the same terms. Generally fixed e...
by billb
Tue Nov 26, 2019 1:57 pm
Forum: runmlwin user forum
Topic: 95% prediction interval for new obs from same cluster populations
Replies: 3
Views: 4327

Re: 95% prediction interval for new obs from same cluster populations

Hi Rachael,
Just generate a level 1 residual (using the level 1 variance) and add it to the prediction and you should be sorted.
Bill.
by billb
Tue Nov 26, 2019 12:40 pm
Forum: runmlwin user forum
Topic: 95% prediction interval for new obs from same cluster populations
Replies: 3
Views: 4327

Re: 95% prediction interval for new obs from same cluster populations

Hi Rachael,
If you were hoping to get 95% intervals for individuals rather than for the average individual you will also need to have level 1 in your predictions I would expect.
Best wishes,
Bill.
by billb
Wed Aug 07, 2019 1:11 pm
Forum: MLwiN user forum
Topic: Wald test and SE of variances
Replies: 2
Views: 4760

Re: Wald test and SE of variances

Hi Sun, the Wald test is only an approximate test for variance parameters as they do not have a normal distribution. That withstanding the test will form the difference of the 2 variances and then work out the SE for this difference to test for significance. This SE is made up of the SE for each par...
by billb
Mon Jul 15, 2019 9:29 am
Forum: MLwiN user forum
Topic: Higher level predictors: are they worth it?
Replies: 2
Views: 5394

Re: Higher level predictors: are they worth it?

Dear John, I would say that it would be important to include the level 2 predictors. In some respect you are largely doing the multilevel modelling to control for variation in the response variable that can be explained by clustering into groups. Of course putting level 2 predictors into the model c...
by billb
Fri Jun 21, 2019 12:32 pm
Forum: runmlwin user forum
Topic: Interpretation of the Raftery-Lewis value
Replies: 2
Views: 10825

Re: Interpretation of the Raftery-Lewis value

Hi Rach, Raftery and Lewis is a completely different diagnostic to the ESS and so works on actual iterations and so you need to run for 19,445 actual iterations to satisfy it. With regard ESS generally one looks for some criterion like > 500 for all parameters. The ESS works on correlations in the c...
by billb
Fri Apr 05, 2019 5:18 pm
Forum: MLwiN user forum
Topic: Measurement Invariance testing with CCMM models
Replies: 4
Views: 6787

Re: Measurement Invariance testing with CCMM models

Hi Pia,
It depends on what model you are fitting. If you look at the MLwiN manuals for MV models you will see that the data is in wide format and then when you specific the columns for the different responses the software constructs the long format in the background.
Best wishes,
Bill.
by billb
Thu Apr 04, 2019 7:16 am
Forum: MLwiN user forum
Topic: Measurement Invariance testing with CCMM models
Replies: 4
Views: 6787

Re: Measurement Invariance testing with CCMM models

Hi Pia, MLwiN has some functionality for multilevel factor analysis (using MCMC) and I suspect it could do cross-classified/multiple membership as well though I have never tested it so I wouldn't vouch for it. I think if you construct factor scores elsewhere and want to fit a more standard multivari...
by billb
Thu Apr 04, 2019 7:05 am
Forum: runmlwin user forum
Topic: Cross-classified logit model: Getting the ICCs
Replies: 8
Views: 13252

Re: Cross-classified logit model: Getting the ICCs

Morning Johannes, I am assuming looking at the syntax you are using runmlwin in Stata? It is probably worth reading a little bit more about how MCMC works e.g. my MCMC in MLwiN book as unlike the IGLS algorithm for ML estimation which converges to ML estimates, MCMC constructs via simulation depende...