-----Original Message-----
From: Rafael Rios Moura [mailto:biorafaelrm at gmail.com]
Sent: Sunday, 11 October, 2020 15:58
To: Viechtbauer, Wolfgang (SP); r-sig-meta-analysis at r-project.org
Subject: Fwd: Overlapping CIs with significant difference among subgroups
ATTACHMENT(S) REMOVED: script_test.R | dataset.csv | pruned_super-tree.tre
Dear Dr. Wolfgang,
I implemented your suggestions and obtained some different results. I found
a difference in the variable field, for example, which represents whether
the data was collected in a laboratory (no) or in the field (yes).
Therefore, the source of the data may introduce a noise in the estimation of
an average effect size. Since the moderator has only two levels, I cannot
add it among random variables. Or is there a way of including field as a
random variable (e.g. effectsizeID|field)? I can also add it as a fixed
predictor. In this situation, how can I estimate the average effect size,
because the intercept will represent one level of the moderator? I added the
variable field to my dataset. Thank you in advance.
Best wishes,
Rafael.