I fitted a mixed effects meta-regression model with one dummy-coded predictor (USA vs. Europe). The output tells me that the predictor is significant. I then wanted to get the average estimated effect size (and CI) for the US and for Europe. I used the predict() function. However when I look at the confidence intervals, they slightly overlap. Why is the p-value of the predictor in the meta-regression model significant, but the confidence intervals of the average effect size in each group overlap?



Browse other questions tagged or ask your own question.