Hi everyone. I am pooling proportions for meta-analysis. As part of this project I have been exploring variance stabilizing transformations for simple proportions and one of the ones that I have looked into is the Freeman-Tukey double arcsine transformation. If x is the numerator and n the denominator of a proportion, the Freeman-Tukey transformation (FT) is given by:

FT = arcsin( sqrt( x / n+1)) + arcsin( sqrt( x+1 / n+1))

The problem is that I have found differing accounts about the variance of this transformation.

Var(FT1) = 1 / (n+1)
Var(FT2) = 1 / (n+0.5)

Which of these is correct?:confused:

PPDoc