# Interpreting meta-regression outputs generated using metafor package

I have been using metafor package to conduct my meta-analysis. I'm quite naive with R software and hence having some difficulty in interpreting the output file generated for meta-regression.

I have used the proportion of females as a moderator variable (continuous data) to assess the effect of gender as a moderator variable. The results of meta-regression shows significant moderating effect of my moderator variable.

My questions are:

1. While reporting the results, do we need to report the value for Q or z?
2. The z-value has a negative sign in mod2 line of Model. What does this sign suggests? I get the exact same results when I use the proportion of males except the z-value is positive. Does this positive or negative z-value suggests the direction of our results in meta-regression?

To answer to your second question, yes, the negative or positive value of the estimate may give you an insight on the direction of the effect of the moderator. Interpretation still dependes on the type of meta-analysis that you are conducting.

To have a graphical representation of this relationship, you may want to plot a so-called "bubble plot". Assuming that you have performed the analysis with the "meta" package, you can produce a bubble plot with a code like this (in the code, "meta.analysis" is the name of the object in which you have stored your meta-analysis model:

metareg1 <- metareg(meta.analysis, mod2)
bubble(metareg1)


The value of QE is a chi-squared statistic for testing the amount of heterogeneity. People often report the value of $$I^2$$ with its confidence interval to give some idea of how precisely the amount of heterogeneity has been estimated. Whether the amount of heterogeneity is of scientific interest depends on your scientific question.

The value of GM is a chi-squared for testing the moderators as a whole. Since you have only one moderator that test is identical to the test for the single moderator given by the $$z$$ as you can confirm by looking at the identical $$p$$-values.

If you reverse the direction of scoring of a moderator then you invert the coefficient.