ML multiple imputation

[shanekav]

Ok, first question...

Firstly, congratulations on Stat-JR. I am still trying to get my head around it, but it seems a remarkably farsighted piece of software.

I have been waiting on the the release particularly as I need to run multiple imputation for missing data on a large ML model. In REALCOM it was so slow as to make it impractical. So far, on looking at the software and looking through the manual I have not been able to work out how to do this. Could you advise me on this?

thanks

Shane

[richardparker]

Hi Shane,

Thanks for your query; we have been developing a template, and eBook, to conduct multiple imputation for 2-level models in Stat-JR; we aim to make this available early in the new year - we'll post a link on this forum, and in a news item on the CMM home page, as soon as it's available.

Best wishes,

Richard

[richardparker]

Hi again,

Further to my last post, we've just uploaded some additional templates on our Stat-JR downloads page (http://www.bristol.ac.uk/cmm/software/statjr/downloads/); the files in the "further multivariate normal and mixed response models" folder, in particular, might be of interest to you in the interim, since they can generate imputed datasets.

Best wishes,

Richard

[shanekav]

Great! Thanks, will have a look and keep an eye out.

Shane

[richardparker]

Following on from this thread, if anyone else is interested in software to perform multiple imputation for 2-level models with missing data, please see the newly-released 2LevelImpute Stat-JR template (and supporting Stat-JR eBook) downloadable from here:

http://www.bristol.ac.uk/cmm/software/s ... evelImpute

[liamsmith]

Thank you for your kind words about Stat-JR!foodle
Regarding your question, Stat-JR currently does not have a built-in multiple imputation function. However, you can use external packages like 'mice', 'Amelia', or 'missForest' in R to perform multiple imputations on your dataset, and then use Stat-JR to fit your ML model on each imputed dataset and combine the results.

[JoseSpencer]

I used the 'Amelia' package. You can try it