Hi, I’m working on a meta-analysis studying different cognitive biases. I’d like to compute the effect size of attention bias index generated from dot-probe task through between-group comparison, i.e. comparing the index between experimental and control group.
The conventional formula in generating the index is: incongruent trials (mean RT) – congruent trials (mean RT). However, most studies reported only the means and SDs of the congruent and incongruent trials for the 2 groups separately.
Is there any way to generate the SD of the index score (besides asking the authors for the original data)? I know we can also derive effect size from t/F/ exact p value, but if these data are not available, then what can be done?
so this index is basically Ma-b=Ma-Mb , where Ma and Mb are collected from a within subject design? if so, you want to find SDa-b ?
there is a formula in this paper (https://pdfs.semanticscholar.org/d88e/7804b7df89d229fd5db7a1c567b94ff18f33.pdf ; formula B1 in the appendix) to estimate this SDa-b . Unfortunately this formula needs you to assume a pretest-posttest correlation value (r), and this r is usually not readily available from the paper. Nevertheless, many meta-analyses have gotten away with simply assuming this r to be .5 .
Or you can calculate the average r from the studies that have provided you enough info to do so, and then use this average r to calculate SDa-b for the other studies.
Thanks! I also read from else where that we can do a sensitivity analysis by assuming r to be 0.1/0.5/ 0.9. I'll see which one is more appropriate, many thanks!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis