[R-meta] Significantly group difference, overlapping confidence intervals

전경남 fstyle71 at naver.com
Thu Feb 1 06:53:06 CET 2018

Dear all,
I conducted a meta analysis using the meta package.
When I did a moderator analysis using a categorial variable(with two levels) as a moderator,
I found a statistically significant moderating effect at the level of .05.
But when I looked at the 95% of confidence intervals of the two pooled effect sizes,
I found there is an overlap between the two confidence intervals.
How I can accurately interpret these two seemingly contradictory results?
Thank you for your advices in advance!
Kyungnam Jeon
* Here I posted the results of my analysis for the moderation effects
Results for subgroups (random effects model):
                            k      SMD            95%-CI             Q       tau^2       I^2
level = ele           10   0.1873   [-0.0018; 0.3764]   40.71   0.0696    77.9%
level = second     5   -0.0538   [-0.1990; 0.0915]     6.99   0.0117    42.8%
Test for subgroup differences (random effects model):
                                Q   d.f.   p-value
Between groups   3.93    1     0.0475

	[[alternative HTML version deleted]]

More information about the R-sig-meta-analysis mailing list