2

I am conducting a meta-analysis and trying to extract standardised differences from a study using pre-post (or repeated measures) design. I am given the Mean, SD, and Sample Size for the pre- and post-test scores (in terms of the metafor package, I have the m1i, m2i, sd1i, and sd2i).

The repeated measures standardised difference takes into account the correlation between the measures (e.g. see a brief description of the method in a previous post here). However, I do not have this information and recommended not to guess it. Are there any ways I can find the repeated measures standardised difference without knowing the correlation between the measures?

(a simple guide on how to do that in R would be sufficient, but feel free to add any mathematical details)

Many thanks!

makie
  • 41
  • 3

1 Answers1

1

Without a value for the correlation or some way of imputing it you cannot compute the standard error of the mean differences. You could always impute a range of values over the plausible range and see how much it affects the analysis but if your advisors are against any form of imputation that is ruled out.

If these are randomised trials then you could just analyse the post scores as the effect of controlling for pre scores should just be to increase precision.

mdewey
  • 16,541
  • 22
  • 30
  • 57