jamovi

mcfanda@gmail.com wrote: I pushed a new master (0.9.7) with the last version of .nicifyTerm() as you suggested. In linux works well, it should be ok in other OS too.

It also works well with windows 7 and 10.

Cheers,
Maurizio

Hi Marcello,

Thanks once again for your software. I have a question regarding your calculation of CI for the multinomial regression. I am getting different 95% CI when I run your module vs. when I run the multinom regression in R and SPSS. I am using multinom function in r.

Thanks,

Hi, thanks for the suggestion. Do you have a working example that I can check? In the meantime, please consider that the way you compute the CI for glm() models may influence the results. GAMLj relies on R emmeans package, which does not necessarely uses the exact same approach of other packages.

To give you a concrete example, I've noticed myself that emmeans CI are different (not drammatically) from confint() function CI in logistic regression. I was puzzled and I asked the emmeans developers. Their swift reply was quite convincing. Maybe we are talking about something similar here. Please check out https://github.com/rvlenth/emmeans/issues/7

Thanks for the link! That might be the case since I used confint() function.

Good, please let us know if you see any other discrepancy

is it possible to add factor loading values into path diagram as in amos?

Hi, it's not clear to me how you would add factor loadinga to a mixed model. If you intend mixed SEM, they are not implemented in gamlj, and it would take a while before we do that. Almost certainly, they will be in a different module.

First, thank you so much for the module!

Would it be possible to also add effect sizes to the mixed models? Perhaps along the lines of Feingold (JCCP 2011, 2015)?

We are working on this. Most likely, we will use effect sizes based on variances, because the d is applicable to a much smaller variety of designs