jamovi vs jasp meta analysis discrepancy troubleshooting

[madelinestein]

hiya,

just wondering if anyone is aware of a known discrepancy between jamovi and jasp meta analysis outputs?

currently working on a project in jamovi with the major package, exported the data into a csv for jasp, and when i re-ran it in jasp, egger's bias test was non-sig (but highly significant in jamovi) and their were minor value discrepancies in heterogeneity, effect size etc.

The discrepancies in values could be a matter of rounding, but unfortunately when i ran fisher's r-to-z data (built-in for jamovi, computed in jasp), the magnitude of the discrepancy increased significantly. Additionally, i don't believe a round error would lead to such a different egger's bias result.

and to further complicate the pot, when doing sub-group analyses (applying filters in both softwares with raw correlations), one (out of four) of the subgroup analyses comes out identical between the two softwares.

Any experience or troubleshooting ideas would be welcome

[MAgojam]

Hey @madelinestein,
if you could provide a file with data producing your findings, i might take a look, before answering.

Cheers,
Maurizio

[madelinestein]

Thanks Maurizio, I got it sorted (re-ran in r which corroborated jamovi results). The simplest conclusion is that there was an error in how i calculated the SE... but I don't think I would have gotten identical results within the subgroup analyses. Mystery for another day.