GAMLj: Very different results after the last update

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)
TPalma
Posts: 24
Joined: Fri Jul 14, 2017 9:04 am

GAMLj: Very different results after the last update

Post by TPalma »

Hi,

I just updated the GAMLj module and after doing so the results of different GLMM analyses I was working on drastically changed. I did not do any changes to the models, just the update. Take a look at the attached files.

Thanks.
Tomás
Attachments
after.png
after.png (103.55 KiB) Viewed 8687 times
before.png
before.png (60.21 KiB) Viewed 8687 times
Last edited by TPalma on Tue Sep 28, 2021 12:27 pm, edited 1 time in total.
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj: Very different results after the last update

Post by mcfanda@gmail.com »

Hi
I cannot see the pictures well. Would you mind posting them again?
TPalma
Posts: 24
Joined: Fri Jul 14, 2017 9:04 am

Re: GAMLj: Very different results after the last update

Post by TPalma »

Done.
As I said, same random structure, same fixed factors (simple contrast coding), same covariate (postdiction, cluster based centered), same method to calculate dfs, different results. Weird.
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj: Very different results after the last update

Post by mcfanda@gmail.com »

Hi Tomas,
I'm afraid I know what is going on. In the previous version, there was a bug in cluster-centering, happening for weired combinations of cluster type and order of cases. The latest version tried to fix it but failed. I've now prepared the update to hopefully fix the issue. This will be version 2.5.2 in a few hours.
For your case, I'd suggest to re-run the analyses with the new version, and check if you get the same as the "before" results.
Furthermore, I've noticed that in your model you have two cluster variables. Keep in mind that "cluster-based centered" centers the variable by the first cluster appearing in the cluster variables list
TPalma
Posts: 24
Joined: Fri Jul 14, 2017 9:04 am

Re: GAMLj: Very different results after the last update

Post by TPalma »

Hi Marcello,

Okay, I'll re-run the analyses with the new version. And thanks for the tip bout cluster-based centered - I did not know that the centering was done by the first cluster variable. I will also check this on the new analyses.

Best,
Tomás
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj: Very different results after the last update

Post by mcfanda@gmail.com »

Centering on more clusters, especially when they are cross-classified, is not that straightforward, so the idea is to leave the user free to prepare the right data beforehand.
TPalma
Posts: 24
Joined: Fri Jul 14, 2017 9:04 am

Re: GAMLj: Very different results after the last update

Post by TPalma »

Hi Marcello,

I just updated to the new version but now I get this error. Can't run the analysis.
Attachments
Screenshot 2021-10-02 at 19.18.23.png
Screenshot 2021-10-02 at 19.18.23.png (169.38 KiB) Viewed 7781 times
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj: Very different results after the last update

Post by mcfanda@gmail.com »

Please update (again) and try again.
TPalma
Posts: 24
Joined: Fri Jul 14, 2017 9:04 am

Re: GAMLj: Very different results after the last update

Post by TPalma »

I have uninstalled it and installed it again but not yet working - see screenshot
Attachments
Screenshot 2021-10-04 at 15.45.44.png
Screenshot 2021-10-04 at 15.45.44.png (180.58 KiB) Viewed 7650 times
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj: Very different results after the last update

Post by mcfanda@gmail.com »

Wow, this bug is really persistent! Let's see if version 2.5.3 solved the issue.
Post Reply