Search found 112 matches

by billb
Wed Dec 12, 2018 11:28 am
Forum: MLwiN user forum
Topic: level 1 variance
Replies: 1
Views: 883

Re: level 1 variance

Hi Megan,
As you are fitting a binary outcome the model is a Binomial model and so is there no level 1 variance i.e. for the binomial distribution the variance is a function of the mean so therefore the box disappears. Hope this makes sense.
Bill.
by billb
Tue Nov 06, 2018 10:24 am
Forum: MLwiN user forum
Topic: Related variables at different levels
Replies: 2
Views: 1957

Re: Related variables at different levels

Hi John, It is quite standard to construct compositional variables in a multilevel model and add them to the model as contextual effects on top of the individual effects. If you think about our example datasets we talk about gender and school gender and one would use the gender variable to construct...
by billb
Tue Nov 06, 2018 10:21 am
Forum: MLwiN user forum
Topic: Heteroscedasticity
Replies: 3
Views: 2070

Re: Heteroscedasticity

Hi Rina,
Presumably you can do a likelihood ratio test between the models with and without heterogeneity and keep those predictors that make a significant difference in the random part?
Best wishes,
Bill.
by billb
Tue Oct 16, 2018 3:47 pm
Forum: MLwiN user forum
Topic: Finite population correction
Replies: 3
Views: 1692

Re: Finite population correction

Hi John,
There isn't a whole lot of literature on FPCs for multilevel models and MLwiN assumes an infinite population in all it's calculations. A search will find a few recent papers if you are interested.
Best wishes,
Bill.
by billb
Mon Oct 15, 2018 1:22 pm
Forum: runmlwin user forum
Topic: Cross-classified logit model: Getting the ICCs
Replies: 8
Views: 3059

Re: Cross-classified logit model: Getting the ICCs

Hi Johannes, I am not sure here whether you mean ICC or VPC but I can offer the paper I wrote on variance partitioning in multilevel logistic models (Browne et al., 2005 JRSS A https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1467-985X.2004.00365.x ). This covers overdispersed models which are...
by billb
Fri Oct 05, 2018 10:23 am
Forum: runmlwin user forum
Topic: Specifying orthogonal in Hierarchical Centring decreases ESS
Replies: 2
Views: 2130

Re: Specifying orthogonal in Hierarchical Centring decreases ESS

Hi KazimovHH, Thanks for the email. Hard to know what the right answer here as you have 2 techniques (Orthogonal parameterisation and Hierarchical centering) which do different things and can both improve mixing for particular model features so the answer will depend on your particular dataset. Orth...
by billb
Fri Aug 10, 2018 4:40 pm
Forum: MLwiN user forum
Topic: Power/Effective Sample Size
Replies: 2
Views: 2142

Re: Power/Effective Sample Size

Hi Divyamore, Just to reiterate my reply to Jillian. MCMC has a concept called effective sample size which has nothing to do with power but is the number of effective independent iterations of an MCMC chain run. This is what MLwiN is reporting (and I should know as one of the programmers!) and altho...
by billb
Mon Aug 06, 2018 7:43 am
Forum: R2MLwiN user forum
Topic: Different ESS in MLwiN and R2MLwiN
Replies: 11
Views: 7319

Re: Different ESS in MLwiN and R2MLwiN

Morning Andy and Chris, When I wrote the ESS calculation in MLwiN 20 odd years ago as you'll see I decided to limit the calculation to the first 1000 elements of the autocorrelation function probably partly because of time/space considerations (computers were slower at the time) and partly because i...
by billb
Mon Jul 30, 2018 8:50 am
Forum: runmlwin user forum
Topic: cross-classified/cross-nested model: How to get all relevant random components
Replies: 1
Views: 827

Re: cross-classified/cross-nested model: How to get all relevant random components

Hi Johannes, Basically in MLwiN (and therefore RunMLwiN) the classifications fitted are the ones you specify - in this case the 4 given each with a set of random effects. If you want to include 'interaction classifications' you need to create the appropriate variables and declare them as additional ...
by billb
Tue May 29, 2018 1:50 pm
Forum: runmlwin user forum
Topic: Can runmlwin handle CAR models at two levels?
Replies: 3
Views: 1596

Re: Can runmlwin handle CAR models at two levels?

Hi Yusuf,
I guess you could and the dependence structure would presumably be easy with each year a neighbour of it's next and last but this is answering something completely different as this would be overall year effects and not different year effects per neighbourhood.
Best wishes,
Bill (no y!)