I am currently in the process of trying to run a multivariate meta-analysis model in metafor with random effects. I have no previous experience working with meta-analysis models and have read many posts/inquiries to try and get my bearings.

Here is the model structure I am currently hoping to proceed with:

acclimation_model<-rma.mv(yi, vi, data=dat_acclim_ES, 
            mods = ~flux_range * mean_temp_reared + exp_age + size + org_level + exposure_temp, 
               random = ~1 |  study_id/ experiment_id/ response_id,

My current questions are:

  1. Is the nested structure correct here (i.e. accounting for collinearity with multiple responses from within the same experiment from within the same study?) I have read different opinions on have a crossed effect for response_id as opposed to nesting within but am not sure.
  2. Is REML the proper estimation metric as opposed to ML when thinking about comparing AIC values? I have read that with REML AIC values are not comparable?
  3. Is it possible to calculate an R2 or I2 statistic to describe the model fit and success at describing heterogeneity? If so, how?
  4. Are there any other metrics I should be reporting to justify the use of this particular model? If so, which ones and are there resources you can point me to?

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.