I have conducted a meta-regression to investigate the change of effect size as a function of age. I used the mean/median of age for each included study as the predictor. I guess this is the common practice and maybe the only way to do the meta-regression. However, I was criticized by the representativeness of the mean when the age range is relatively wide, such as the mean 40 for a range from 20 to 60. I am wondering whether there are any recognized solutions for such a problem, and if yes, their references. And if possible, how to achieve that with metafor or other packages in R. I can think of a possible solution, that is, to adjust the variance of response variable by taking the variance of predictor into account. In other words, when the age range is wider, the variance of predictor variable is higher. However, after a second thought, this may be problematic, considering (1) How to mathmatically adjust the variance in response varuable? (2) The variance of predictor should influence the certainty of the predictor itself, not the response. I appreciate any suggestions and discussions to this matter.
I have searched the internet but found no useful solution. I expect to get a conceptual and/or practical solution, or at least some hints how I can address the problem.