Pre-post design

Welcome to the forum for R2MLwiN users. Feel free to post your question about R2MLwiN here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Go to R2MLwiN: Running MLwiN from within R >>
Post Reply
Posts: 1
Joined: Mon Jan 06, 2020 12:50 pm

Pre-post design

Post by KenMLM2020 »

Hi everyone,

I'm currently implementing a model for a binary variable of mental health condition (i.e. whether or not someone has the condition)

I have two options in R,one MLM and one standard binomial regresssion, but I'm wanting to check if a) this makes sense the way I'm doing it, 2) if MLwiN would provide improvements in performance as it's currently very slow:

The model is specified like this

MeanLM<- glm(HealthCondition~ XMean+Age+Sex+Diet+Deprivation, data = data, family = binomial)

DayMLMMean <- glmer(HealthCondition~ XMean+Age+Sex+Diet+Deprivation+(0+Age|ID)+(1|ID), data = data,
family = binomial(link=logit))

XMean and Age continuous variables, and the rest are categorical.

The data are specified in long form, with two measures of HealthCondition corresponding to the two difference participant ages at mearsurement before and after exposure (XMean). As it stands I get very similar results for both the simple binomial regression and the multi-level model - which I think makes sense as the data are restricted to two measurements.

I've read the Bristol MLM course on pre-post design (specifically the repeat mesures part). Any tips would be very well received!


Posts: 147
Joined: Fri May 21, 2010 1:21 pm

Re: Pre-post design

Post by billb »

Hi Ken,
Apologies that this one hasn't been answered after so long. Basically I think R2MLwiN has a very similar syntax to glmer and thus will likely fit the same models. They are using different algorithms so I am not sure which will be the quicker. I suspect estimates for 2 models would be similar but larger standard errors in the multilevel model.
Best wishes,
Post Reply