Interaction terms

Welcome to the forum for Stat-JR users. Feel free to post your question about Stat-JR 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!

We will add further support to the Stat-JR website, such as FAQs and tutorials, as soon as it is available; the Stat-JR website can be found here: http://www.bristol.ac.uk/cmm/software/statjr/
Post Reply
shakespeare
Posts: 70
Joined: Thu Feb 14, 2013 11:12 pm

Interaction terms

Post by shakespeare » Tue Jul 22, 2014 3:32 pm

Just want to confirm: When using the 2LevelImpute template in TREE, I see no direct way to enter interaction terms, correct? These terms would need to be added to the data set before reading into TREE for processing, right? Thx.

richardparker
Posts: 57
Joined: Fri Oct 23, 2009 1:49 pm

Re: Interaction terms

Post by richardparker » Tue Jul 22, 2014 4:29 pm

Hi - yes, at the moment interaction terms would have to be generated in the dataset prior to its use with the 2LevelImpute template; they then could be added as another variable, but note that this would introduce bias. Some of my colleagues here at the CMM have recently published on this topic (see below), and we hope to refine the manner in which interactions are handled in future template developments:

Goldstein, H., Carpenter, J. R. and Browne, W. J. (2013), Fitting multilevel multivariate models with missing data in responses and covariates that may include interactions and non-linear terms. Journal of the Royal Statistical Society: Series A (Statistics in Society). doi: 10.1111/rssa.12022 (also downloadable from here: http://www.bristol.ac.uk/cmm/team/hg/fu ... ractns.pdf).

Best wishes,

Richard

shakespeare
Posts: 70
Joined: Thu Feb 14, 2013 11:12 pm

Re: Interaction terms

Post by shakespeare » Tue Jul 22, 2014 5:20 pm

My plan was to follow the suggestion of von Hippel, but understand this is less than optimal. Stef van Buren has worked out an algorithm for squares (http://www.stefvanbuuren.nl/publication ... 20-SMR.pdf), but conclude the transform and impute procedure is the best available approach for interactions between different variables. Clearly that is not as theoretically sound as your colleagues at Bristol, but until the new template is ready, it's the next best choice.

Post Reply