Skip to content

[R-meta] Dealing with missing data in bivariate analysis

3 messages · Olina Ngwenya, Wolfgang Viechtbauer

#
Dear R-sig-meta-analysts

I have been doing bivariate meta-analysis using rma.mv(), but one of my outcomes has missing values and I am getting this message "Rows with NAs omitted from model fitting". My question is "Do we have other ways of dealing with missing data in meta-analysis instead of discarding rows with missing values".

Thank you

Olina Ngwenya
Research Assistant
Centre for Biostatistics | School of Health Sciences | Faculty of Biology, Medicine and Health | University of Manchester
#
Dear Olina,

Depends a bit on what is missing. If values of predictor/moderator variables are missing, then one possibility is to use some kind of imputation technique. See, for example:

https://www.metafor-project.org/doku.php/tips:multiple_imputation_with_mice_and_metafor

Best,
Wolfgang
#
Thank you Wolfgang.

Regards

Olina

-----Original Message-----
From: Viechtbauer, Wolfgang (SP) <wolfgang.viechtbauer at maastrichtuniversity.nl> 
Sent: 28 March 2022 11:06
To: Olina Ngwenya <olina.ngwenya at manchester.ac.uk>; r-sig-meta-analysis at r-project.org
Subject: RE: Dealing with missing data in bivariate analysis

Dear Olina,

Depends a bit on what is missing. If values of predictor/moderator variables are missing, then one possibility is to use some kind of imputation technique. See, for example:

https://www.metafor-project.org/doku.php/tips:multiple_imputation_with_mice_and_metafor

Best,
Wolfgang