Skip to content

[R-meta] effect sizes for pre-post mean differences or within-subject design comparison without correlation info

2 messages · Zhang Yikang, Wolfgang Viechtbauer

#
Dear r-sig members,

I am trying to conduct a meta-analysis examining how social feedback impact people?s belief in their memories. Normally the design of the studies include ratings of their memories before and after the social feedback (i.e., a pre-post comparison) or a within-subject design where some memories receive feedback and others don?t.
In both situations, the ratings between the two groups are correlated. However, to calculate the effect sizes of such design, metaphor also needs the input of the correlations between the  pre and post ratings (or ratings in control  and feedback conditions), which are unavailable to me.

So, I was wondering if there is a way to by-pass the stats of correlations and just use mean and SD.  And if there is none, is it justified to treat the data as if they were independent and compute the effect sizes using the ?SMD? method and acknowledge this comprise in the manuscript?

Many thanks for your help!
Best regards,
Yikang

Yikang Zhang|???, PhD Student
Faculty of Psychology and Neuroscience (Forensic Psychology Section<https://www.flpmaastricht.org/>), Maastricht University
UNS 40 Room 3.749 (TEL 84075)
Lab: Criminological and Experimental Legal Psychology Lab (CELL)<https://www.celleuven.com/> at KU Leuven
3 days later
#
Dear Yikang,

The pre-post correlations are really needed. If you don't know them, then you could use 'guestimates'. This has been discussed previously on this list, so search the archives for related discussions. Treating the data as if they are SMDs would not be correct.

Best,
Wolfgang