Search found 1246 matches

by ChrisCharlton
Wed May 19, 2021 2:00 pm
Forum: runmlwin user forum
Topic: Area Under Curve (AUC) after random Effects models
Replies: 1
Views: 1025

Re: Area Under Curve (AUC) after random Effects models

I checked this with George, and his advice was to firstly calculate the predicted probability in the usual way and then to use the roctab command to obtain the AUC from this.
by ChrisCharlton
Wed May 19, 2021 11:47 am
Forum: R2MLwiN user forum
Topic: Issues when storing residual chains
Replies: 2
Views: 1072

Re: Issues when storing residual chains

Would it be possible for you to confirm that the following, which should be an equivalent model using the MLwiN example data works for you? > library(R2MLwiN) > data(bang, package = "R2MLwiN") > f <- logit(use) ~ 1 + age + (1 | district) > m <- runMLwiN(f, D = "Binomial", data = ...
by ChrisCharlton
Tue May 11, 2021 5:29 pm
Forum: runmlwin user forum
Topic: Predict after runmlwin
Replies: 1
Views: 804

Re: Predict after runmlwin

I think that you have an extra comma that shouldn't be there in the:

Code: Select all

level2(Period:cons, , residuals(v))
part of your syntax. Otherwise what you wrote should work. See chapter 3 of the MLwiN manual replication materials on https://www.bristol.ac.uk/cmm/software/ ... /examples/ for an example.
by ChrisCharlton
Tue May 11, 2021 5:21 pm
Forum: runmlwin user forum
Topic: Combining mcmc(cc) logit results using mi estimate
Replies: 1
Views: 897

Re: Combining mcmc(cc) logit results using mi estimate

I asked George about this and he had the following suggestions, although we are not that familiar with the mi estimate command: You can probably skip step 1 and just supply some plausible starting values to step 2, rather than having to do two sets of imputations. Otherwise the runmlwin syntax looks...
by ChrisCharlton
Tue May 04, 2021 12:29 pm
Forum: MLwiN user forum
Topic: Good-of-fit measure in poisson multilevel model
Replies: 1
Views: 810

Re: Good-of-fit measure in poisson multilevel model

If you're fitting with MCMC then you could compare the DIC statistics, otherwise we would normally suggest doing a Wald test.
by ChrisCharlton
Mon Apr 26, 2021 10:27 am
Forum: MLwiN user forum
Topic: Set up of equations
Replies: 1
Views: 729

Re: Set up of equations

If you open the Model->Hierarchy Viewer and click the Options button you can specify a 0/1 column with which to exclude rows from the model. This corresponds with the EXCL command.
by ChrisCharlton
Mon Apr 26, 2021 9:10 am
Forum: MLwiN user forum
Topic: Installing MLwiN on Mac
Replies: 1
Views: 729

Re: Installing MLwiN on Mac

The following page describes a variety of options for running the software on a Mac:
https://www.bristol.ac.uk/cmm/software/ ... ysreq.html
by ChrisCharlton
Tue Apr 20, 2021 6:41 pm
Forum: runmlwin user forum
Topic: Average Marginal Effects via runmlwin
Replies: 2
Views: 1013

Re: Average Marginal Effects via runmlwin

My interpretation of this is that he is suggesting that you duplicate each row of your dataset 1000 times (presumably with https://www.stata.com/help.cgi?expand ) so that you have a copy for each MCMC iteration. You would then combine this with the parameter estimates for the corresponding iteration...
by ChrisCharlton
Fri Apr 09, 2021 3:43 pm
Forum: R2MLwiN user forum
Topic: Interaction plots for mlwinfitMCMC fixed effects
Replies: 2
Views: 1121

Re: Interaction plots for mlwinfitMCMC fixed effects

You could manually create the data required for this plot by generating a chain of predictions and then calculating the mean and credible intervals from this. An example of this would be as follows: library(R2MLwiN) library(coda) data(tutorial, package = "R2MLwiN") (model1 <- runMLwiN(norm...
by ChrisCharlton
Thu Apr 08, 2021 3:58 pm
Forum: MLwiN user forum
Topic: Input columns have unequal length
Replies: 2
Views: 961

Re: Input columns have unequal length

I suspect that the command is complaining that your 'CNTSCHID' column is a different length to the others. All the input columns involved in this command need to be of the same length. c1500 is included in the command as this is temporary output column where it will put the short version of the ID c...