Skip to content

[R-meta] Meta-analysis on positive and negative outcomes: Is it correct to take the reversed r values?

3 messages · Gladys Barragan-Jason, Wolfgang Viechtbauer

#
Dear all,
I have a question about taking the reversed values when combining different
proxies of one factor. For instance, if you're doing  a meta-analysis on
the impact of X on Y where Y is  mental health and you have several
measures of Y such as depression (r= -0.4) , stress (r=-0.1) and emotional
balance (r=+0.5-) and positive moods (r=+0.2), is it correct to transform
(take to reversed values) for "negative health" (r=+0.4 and r=0.1) and
pooled everything together or is it better to make to different
meta-analysis (one on positive and one on negative effects), or maybe both?
Thanks a lot for your response.
Best,
Gladys
#
Dear Gladys,

Yes, you can flip the sign. In essence, this is what would happen if a study where Y reflects depression had reverse-scored their depression measure (such that higher values reflect *less* depression).

Whether the association between X and 'positive' measures of mental health is the same as the association between X and 'negative' measures of mental health (after flipping the sign of the latter) is a different issue. With enough studies, one could code a moderator variable to indicate whether Y was originally a positive or negative measure and examine whether the association between X and Y differs across these two groups.

Best,
Wolfgang
#
Thanks a lot Wolfgang.
That is very helpful.
Best,
Gladys

Le mar. 19 janv. 2021 ? 09:31, Viechtbauer, Wolfgang (SP) <
wolfgang.viechtbauer at maastrichtuniversity.nl> a ?crit :