In a random-effects meta-analysis I found a negative association between a certain personality trait and gratitude (higher score on personality scale is associated with lower gratitude). Pearson's correlation coefficient is my effect size.

This association is significantly moderated by age. The estimate in the moderator analysis is negative.

Does this mean that the negative association between personality and gratitude gets more negative (i.e., stronger) with increasing age? Or is it getting less negative (i.e., weaker)?

The analysis was done using the robumeta package in R.

  • $\begingroup$ I would interpret that as the higher the age, the lower the gratitute, independently from personality score $\endgroup$ Apr 23, 2018 at 9:13
  • $\begingroup$ What effects size are you using? Mean differences? Correlations? Perhaps if you gave more detail by editing your question the answer might become clearer. $\endgroup$
    – mdewey
    Apr 23, 2018 at 17:14
  • $\begingroup$ I edited my original post. Pearson's correlation coefficient was used as effect size. $\endgroup$
    – Marie B.
    May 4, 2018 at 9:43
  • $\begingroup$ If you could paste your output summary for moderator analysis., this may help us to answer your problem correctly. $\endgroup$
    – user10619
    May 5, 2018 at 9:31
  • $\begingroup$ Given the description, you seem to be working on point- biserial correlation - effect size for your meta- analysis. $\endgroup$
    – user10619
    May 5, 2018 at 9:37

1 Answer 1


If you are analysing correlation coefficients and the summary estimate is negative then a moderator with a negative coefficient corresponds to the statement that studies with a high value of the moderator (in this case age) have more negative correlations. You are using the robumeta package so you are doing a meta-analysis of summary statistics. This is, therefore, an ecological finding since you appear to be using mean age (or some other summary of age like the median) in the study as the moderator so it does not mean that older people have more negative correlations but that studies with on average older people tend to have more negative correlations.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.