Skip to content

[R-meta] metafor: standardized mean change using raw score standardization

3 messages · Spyridon Siafis, Wolfgang Viechtbauer

Spyridon Siafis
# 
Hello to all,

I would like to calculate standardized mean change with raw score standardization ("SMCR") using the escalc().
Therefore: sd1i=pre-test standard deviation, m1i=  mean change score or post-test mean and m2i=0 or pre-test mean, respectively.

As far as I understood, the arguments sd2i and ri are not needed for the calculations, but they are required in escalc.
Could I use sd2i=0 and ri=0? Do they influence the calculations?

Thank you very much in advance!

All the best,
Spyros Siafis
Wolfgang Viechtbauer
# 
Dear Spyros,

Indeed, 'sd2i' isn't needed, since escalc() computes yi=(m1i-m2i)/sd1i for 
measure="SMCR". If you have the mean change (and not the pre- and 
post-test means), then you can set m1i to the mean change and m2i to 0. 
But you will need to specify 'ri' and cannot simply assume it to be 0. The 
value of 'ri' won't change the size of 'yi', but it will influence the 
size of the sampling variance ('vi').

Best,
Wolfgang
Spyridon Siafis
# 
Dear Wolfgang, 

Thank you very much for your clarification!
It was not so clear to me.

Best regards,
Spyros

-----Urspr?ngliche Nachricht-----
Von: Wolfgang Viechtbauer <wolfgang.viechtbauer at maastrichtuniversity.nl> 
Gesendet: Wednesday, August 7, 2019 2:05 PM
An: Spyridon Siafis <spyridon.siafis at tum.de>
Cc: r-sig-meta-analysis at r-project.org
Betreff: Re: [R-meta] metafor: standardized mean change using raw score standardization

Dear Spyros,

Indeed, 'sd2i' isn't needed, since escalc() computes yi=(m1i-m2i)/sd1i for measure="SMCR". If you have the mean change (and not the pre- and post-test means), then you can set m1i to the mean change and m2i to 0. 
But you will need to specify 'ri' and cannot simply assume it to be 0. The value of 'ri' won't change the size of 'yi', but it will influence the size of the sampling variance ('vi').

Best,
Wolfgang