Although I plan to use moderators to understand how the raw mean (and its sampling variance) is related to specific experimental conditions, I am curious: how would you interpret the intercept-only model results when the effect size represents a distribution of values from a single group rather than a difference between groups (e.g., when using raw mean difference or Hedges' g)?

For instance, in R{metafor}, I could calculate the sampling variance for response x

play = escalc(measure = "MN", mi=mx, sdi=sdx, ni=nx, data=play)

And then run a simple intercept-only model

mod = rma(yi, vi, data = play)

What does it mean if the intercept is significant? Does it mean that the raw mean effect size does not cross (i.e., is different from) zero? And is the intercept value an indication of the "real" (i.e., sample variance-adjusted) mean of the response?


1 Answer 1


You are analyzing raw means, so in a random-effects model, the intercept represents the estimated average (over studies) mean (so, assuming there is heterogeneity, the true means differ across studies and the intercept term is then the estimated average mean). Therefore, if the intercept is significantly different from zero, then this suggests that the average mean is unlikely to be 0.

  • $\begingroup$ Thank you. You are always very helpful, and I appreciate it. $\endgroup$
    – chabeck
    Jan 20, 2016 at 0:34

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.