1
$\begingroup$

Does the REML estimator (default) in the R metafor package provides inverse-variance weighting?

$\endgroup$
4
  • 1
    $\begingroup$ Are you asking (a) if it is possible to do inverse-variance weighting with REML in the metafor package or (b) does the REML estimator do this by default? $\endgroup$ – Mark White Jul 20 '17 at 20:33
  • $\begingroup$ first b, if not a + how $\endgroup$ – Nadine Jul 20 '17 at 20:41
  • $\begingroup$ What code are you using to run your model? rma(...)? $\endgroup$ – Mark White Jul 20 '17 at 20:48
  • $\begingroup$ For future questions about R code which often get closed here you might like to try the recently established mailing list for R meta-analysis queries stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis You do need to register before posting $\endgroup$ – mdewey Jul 21 '17 at 12:56
2
$\begingroup$

The documentation for the metafor package lays out how it does the weighting for both fixed- and random-effects models. Random-effects models will use weights that are—in part—calculated using the inverse of the variance. However, it also includes an estimate of $\tau^2$ in the denominator (added to the variance). This is specified on page six of the documentation I linked to above. There are various types of estimation methods you can use, and they don't involve changing how weights are determined. Different types of weighting (metafor lets you define your own weights, if you want) can be used with different types of estimation.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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