Hello,
I am currently running meta-regression models using 2-3 variables. I have coded my categorical variables and included them with my continuous variables in analysis. Am I correct to assume that running the regression with the variables this way would be the same as creating a subgroup analysis for my categorical variable? Is there any reason to do this subgroup analysis rather than simply run my regression with the variables dummy coded? I am using Comprehensive Meta-Analysis software and only have one variable that is categorical.
A univariate subgroup analysis is indeed mathematically equivalent to a meta-regression model with a single categorical predictor. There is no reason to prefer the one over the other, unless you are interested in testing your 2-3 explanatory variables simultaneously. This would be done using a multiple meta-regression model, which allows you to test for subgroup differences while controlling for potential confounding effects of your remaining explanatory variables.
Thank you for your reply! And what if I were to enter 2-3 variables simultaneously without subgroup analysis - would this still be valuable information? am I correct to understand that cma treats the variables as moderators? woukd they control for each other or take each other into account?
Subgroup analysis investigates whether the results are the same for each group. It is preferable to use continuous data as continuous values. To use categorical data in regression analysis, the fit of the regression is poor unless there are 10 data in the smallest category. If you use multiple variables, check the correlation between them. If the correlation is strong, it is best to select a representative one.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence