[R-meta] Subgroup comparison for non-linear meta-regression

[Viechtbauer Wolfgang (SP)]

Dear Mustafa,

Subgrouping and then comparing the estimates across subgroups is in no way more efficient than running the model with the interaction. In fact, if one runs the model with the interaction and allow the amount of heterogeneity to differ across subgroups, then the two approaches are identical. See here:

http://www.metafor-project.org/doku.php/tips:comp_two_independent_estimates

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Mustafa Kilickap
Sent: Thursday, 26 October, 2017 14:45
To: r-sig-meta-analysis at r-project.org
Subject: [R-meta] Subgroup comparison for non-linear meta-regression

Dear all,

I did a “non-linear” meta-regression for prevalence (proportion) on time variable (year) using a power function. The studies consist of information for male and female participants. I would like to assess
whether the non-linear fit was similar for male and female. In other words, I would like to

make a comparison for male and female. 

I think it is more appropriate to do a meta-regression including gender*time interaction, however, number of the studies is small. My questions are:

1) Is it acceptable just to compare the subgroups (as stated below) without using interaction term. If it is acceptable;

2)    Does metafor package has a command like “byvar” that makes a statistical comparison of two subgroups possible? If not,

2)    Is it appropriate to calculate z=(p1-p1)/sqrt(se1^2 +se2^2) and then p for a “non-linear” relationship?
 
Kind regards
Mustafa Kilickap