Disappearing interaction term in GAMLj linear models

Everything related to the development of modules in jamovi
Post Reply
richlopez41
Posts: 3
Joined: Mon Feb 19, 2018 6:48 pm

Disappearing interaction term in GAMLj linear models

Post by richlopez41 »

Hi there, I've come across something strange when using the GAMLj module in jamovi, specifically when I run a model with X1, X2, and X1*X2 (their interaction) as predictors.

After the model has been run and the output is saved and appears on the right, if I click *back* in to check options or edit model specifications, the interaction term disappears from the "Model Terms" field, and I have to re-specify/compute the interaction term so it re-appears in the results again.

I realize this is a small inconvenience, but if anything can be done in future releases of GAMLj to remedy this, that would be great!

Thanks,
Rich
dropmann
Posts: 79
Joined: Thu Feb 02, 2017 10:26 am

Re: Disappearing interaction term in GAMLj linear models

Post by dropmann »

Hi richlopez41,

Even though this is Marcello's module I have had a look to see if the issue lies outside the module. I can't, however, reproduce the issue. Would it be possible to provide a .omv file in which it occurs, or a little screen grab video of it happening? Also you did provide details of how you achieved this issue, however, could you also write down the step by step process from opening jamovi, this would be helpful in us trying to pin it down.

thanks
dropmann
Posts: 79
Joined: Thu Feb 02, 2017 10:26 am

Re: Disappearing interaction term in GAMLj linear models

Post by dropmann »

It would also be good to submit an issue to the GAMLj github repo at https://github.com/mcfanda/gamlj/issues. :)
User avatar
mcfanda@gmail.com
Posts: 452
Joined: Thu Mar 23, 2017 9:24 pm

Re: Disappearing interaction term in GAMLj linear models

Post by mcfanda@gmail.com »

Hi,
if I understood the issue, it happens when you have interactions with continuous variables. Right? This is because GAMLj follows the (somehow unjustified) SPSS rule of setting the interaction automatically for categorical variables but not for continuous independent variables. This can be fixed quickly.
More complex is the issue that after saving and going back, the module does not remember the specified model. This indeed lies outside the module. We'll look at the issue soon, thanks
richlopez41
Posts: 3
Joined: Mon Feb 19, 2018 6:48 pm

Re: Disappearing interaction term in GAMLj linear models

Post by richlopez41 »

@mcfanda, yes that's correct. You have to initially specify the interaction and that only happens with an interaction term is comprised of continuous predictors (interaction of two factors is automatically specified/computed).

And no rush on the other issue of going back to re-compute the interaction term..if you do figure it out let me know and/or I'll keep an eye out for GAMLj updates.

Thanks for all your hard work in developing this module. It's been incredibly helpful for my various analysis needs!

Best,
Rich
(email: richard.lopez@rice.edu)
Post Reply