Dear SDM fellows,

I am doing a meta-analysis about longitudinal changes of neuroimaging in schizophrenia. Now I have some questions on effect size estimate:

1. If an article reported a significant group×time effect using an repeated measures ANOVA or ANCOVA based on a flexible factorial design in SPM's "specify 2nd level" , how could I convert the Z value or P value in to T value? Should I choose "Two-sample p or z to t converter" with "Size of the first sample" filling the number of  the case group and "Size of the second sample" the control group?

2.If an article reported a significant "within group" time effect using an repeated measures ANOVA or ANCOVA based on a flexible factorial design in SPM's "specify 2nd level" , how could I convert the Z value or P value in to T value? Should I choose "Two-sample p or z to t converter" with "Size of the first sample" filling the number of the case group and "Size of the second sample" also the case group? 

3.If an article reported a significant effect using a paired-t test within the case group, should I simply use the "One-sample p or z to t converter" or "Two-sample p or z to t converter" with "Size of the first sample" filling the number of the case group and "Size of the second sample" also the case group?

I am rather confused. Thank you in advance

Best regards,
Chengyu

Dear Chengyu

you should choose "two-sample" when the contrast compares the sample of interest (e.g. patients) with a reference sample (e.g. controls), whereas you should choose "one-sample" when the contrast only applies to one sample (e.g. pre-post differences).

Thus, I guess that in the second and third examples you should choose "one-sample".

I'd like to note that it's normal that the studies included in a meta-analysis have methodological differences, but they should all assess the same effect, e.g., they should only focus on one group, or they should only focus on the comparison between that group and controls.

Best,

Joaquim

Thank you for your answer!