MAJOR "Fisher scoring algorithm did not converge"

[adipa]

I am trying to a do a meta-analysis with MAJOR package using the Effect size and Variance and I am getting this error, it suggests to adjust step length of the algorithm which also does not help. I have used the same data in CMA and RevMan and they give proper results, but now I have to expand the data and am checking Jamovi. Kindly help.

[MAgojam]

Hey @adipa,
if you are using MAJOR v1.2.3 you should have an option to intervene on the step of the Fisher scaling algorithm. Generally the 100 predefined iterations for the algorithm in the rma() function under the hood are enough to achieve convergence.
It may happen that this does not happen and a first suggestion is to intervene on the algorithm step (reducing it).
Last week I updated the Module with this possibility. Let me know and if possible attach a omv with your ES and Sampling Variances for a check.

Cheers,
Maurizio

[SergeyYegorov]

Hi @MAgojam,
I am encountering the same error message when using ORs as Effect size and SEs. Is there a workaround for this?
I know that in Metafor I would adjust the rma function, but is there a way to do this in jamovi?
Thanks!
Sergey

here are my data:
OR_calculated
0.2
0.69
0.255
0.29

SE_calculated
0.234693878
0.12244898
0.113265306
0.402295918

[MAgojam]

Hey @SergeyYegorov,
are you doing this (screenshot) with version 1.2.4 of MAJOR?

ScreenShot_20230523213355.jpeg (330.62 KiB) Viewed 27697 times

Cheers,
Maurizio

[cynthiahs]

Hello, I'm having the same issue, and I enter the data just like in the screenshot shown above. I'm using the jamovi 2.3.28 version. Any suggestions on how to fix this issue? I also tried different Step lengths and I get the same error message.
Thanks in advance!

[MAgojam]

Hey @Cynthias,
if there are no problems sharing the file where the error occurs, I can take a look at it, because your reference to the previous post seems to work with the jamovi solid v2.3.8 and the MAJOR module in the library (v1. 2.4), as in the attached screenshot.

ScreenShot_20230929135852.jpeg (360.35 KiB) Viewed 27235 times

Cheers,
Maurizio

[cynthiahs]

Hi Maurizio, thank you for your reply. I tried to send the file to you but it won't attach to the message.

[MAgojam]

I tried to send the file to you but it won't attach to the message.

Thanks @Cynthias,
but you should probably zip it, do you want to try again?

Cheers,
Maurizio

[cynthiahs]

Thank you @Maurizio. I sent you a private message.

[Yeld]

Hi

Starting in the most recent version of JAMOVI there seems to be a bug in MAJOR -> Effect size and (sampling variances and standard errors). When moderation is selected (either moderator types) the algorithm almost always states that "Fisher Scoring algorithm did not converge" . Does not change when altering step length. Datasets that worked before are now no longer working.

Can anyone help?