Search found 424 matches

by GeorgeLeckie
Fri May 12, 2017 4:18 pm
Forum: MLwiN user forum
Topic: Question about ICC
Replies: 1
Views: 514

Re: Question about ICC

Dear Sherry08, If you have found a neighbouhood level ICC of 0.96 for BMI, then I suggest something has gone wrong in your calculation. I would expect something more like 0.05! The ICC is the correlation between two randomly selected individuals' BMI from the same neighbourhood and so we would never...
by GeorgeLeckie
Thu Mar 23, 2017 2:27 pm
Forum: runmlwin user forum
Topic: stata imputed data in mlwin
Replies: 4
Views: 2065

Re: stata imputed data in mlwin

Dear Rodrigo, Suggest you run the mo estimate command with the noisily option and inspect each set of model results. It may well be the case that one of your variance components gets stuck on zero on one or more iterations. mi estimate then complains as it assumes that the parameter stuck on zero ha...
by GeorgeLeckie
Tue Mar 07, 2017 6:33 pm
Forum: MLwiN user forum
Topic: homoscedasticity assumption in logistic models?
Replies: 2
Views: 1156

Re: homoscedasticity assumption in logistic models?

Dear adeldaoud, Yes there is no level-1 residual in the linear predictor of a two-level random-intercept logistic regression and so we only assess the distributional assumption of the level-2 random effect, namely that it is normally distributed and homoskedastic. You can assess the plausibility of ...
by GeorgeLeckie
Thu Feb 23, 2017 5:08 pm
Forum: MLwiN user forum
Topic: Problem with correlation between intercept and slope
Replies: 1
Views: 715

Re: Problem with correlation between intercept and slope

Hi Max, The correct correlation to report is the one based on the estimated variances and co-variances. The correlation based on the predicted random effects differs in part due to the fact these are shrinkage estimates. You should find some coverage of this in various multilevel text books. As an a...
by GeorgeLeckie
Sun Feb 19, 2017 1:31 pm
Forum: runmlwin user forum
Topic: coefficients and variance partition differs considerably between runmlwin and melogit
Replies: 2
Views: 1157

Re: coefficients and variance partition differs considerably between runmlwin and melogit

Dear Lisa, The differences you see relate to the severe clustering exhibited in your data. MLwiN's default estimation method for binary and other categorical and count response models is quasilikelihood estimation and comes in four flavours of increasing accuracy, MQL1 (default), MQL2, PQL1 and PQL2...
by GeorgeLeckie
Tue Jan 24, 2017 10:07 am
Forum: runmlwin user forum
Topic: Use of predict and margins post-estimation
Replies: 1
Views: 1249

Re: Use of predict and margins post-estimation

Hi Marie,

I'm afraid you will have to make your graphs from first principles. We have not done anything in runmlwin to allow margins to work with it; it is a non-trivial task.

Best wishes

George
by GeorgeLeckie
Fri Jan 06, 2017 4:36 pm
Forum: MLwiN user forum
Topic: interpreting main effect in random slopes model
Replies: 1
Views: 850

Re: interpreting main effect in random slopes model

Dear emazel90, I would keep the time trend in the fixed-part of the model even if the slope is not significant. The fixed-part time trend is such a fundamental part of the model, it would look odd to remove it, even though it is not significant. Essentially the population mean does not change over t...
by GeorgeLeckie
Thu Sep 15, 2016 10:51 am
Forum: runmlwin user forum
Topic: Multilevel piecewise logistic regression
Replies: 3
Views: 1230

Re: Multilevel piecewise logistic regression

Hi IgnacioA, I am not entirely sure what you are after. The question also seems more about working our what you want to do in terms of specifying your underlying statistical model rather than how to implement a given model in runmlwin. What might help is to first consider a standard piecewise linear...
by GeorgeLeckie
Fri Sep 02, 2016 10:53 am
Forum: runmlwin user forum
Topic: Testing variance components across equations
Replies: 2
Views: 1282

Re: Testing variance components across equations

Hi Phillipe, Wald tests are frowned upon when testing whether variance components are significant as variances do not have normal sampling distributions. For this reason, likelihood ratio tests are generally preferred in this setting. In your case you are interested in testing whether the difference...
by GeorgeLeckie
Fri Sep 02, 2016 10:49 am
Forum: MLwiN user forum
Topic: Calculating CI variables in multi level multiple regression models
Replies: 2
Views: 3649

Re: Calculating CI variables in multi level multiple regression models

Hi Helen,

In the latest version of MLwiN, the Model Comparison window now gives you p-values and 95% confidence intervals for each regression coefficient if you tick the relevant check boxes.

Best wishes

George