[R-meta] Moderators: controlling for within and between effects

[Farzad Keyhan]

Dear All,

I'm doing a meta on the factors influencing the effect of professional
development (pd).

Each effect estimate in each study may come from: (A) questionnaires
with different number of items [`items`], (B) different amount of pd
training in hours [`training_hr`], and (C) different amounts of time
elapsed up to the point each effect was measured at in weeks
[`time_wk`].

For each focal, study-level, categorical moderator (e.g., format of
the pd [`pd_format`]), I want to control for A, B, and C across
studies.

Because A, B, and C can vary both within and between studies, I wonder
whether I should control for both their within and between
contributions (fit1) or not (fit2)?

fit1 <- rma.mv(yi ~ 0 + pd_format + time_wk_whn + time_wk_btw +
items_whn + items_btw + training_hr_whn + training_hr_btw, random =
~1|study/obs)

fit2 <- rma.mv(yi ~ 0 + pd_format + time_wk + items + training_hr,
random = ~1|study/obs)

Thank you,
Fred