GLMM - Need some help setting up my data

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)
Post Reply
Rageofanarchy
Posts: 2
Joined: Fri Feb 03, 2023 2:42 am

GLMM - Need some help setting up my data

Post by Rageofanarchy »

Hello there,

I'm a PhD student and have found myself here as I think I need to run a GLMM on the data I collected for a study. I've learned so much about statistics in the last week or so and have gone down the rabbit hole of GLMM (hence why I found Jamovi), but I'm not confident in what my data should look like in terms of set up and how to interpret the results.

I'll outline my study briefly and how I'm trying to set it up in Jamovi - I've had some loading errors which makes me think I'm asking it to calculate too many variables at once, but I'm not really sure what is happening in the background to know definitively.

Ok, so I took four classes of English second language leaners and taught them 20 words with two different methods. The participants in the four classes were divided into Group A and Group B. The former group did words 1-10 with method #1 and words 11-20 with method #2, and the latter group did words 1-10 with method #2 and words 11-20 with method #1.

The data points that I have are:

Participant IDs (39 participants in total)
Pre / Post / Delayed Post tests
The 20 words
What method they used to study each word
Whether they were correct or incorrect on each item (the binary dependent variable)
Group A or B
Class (there were four levels)

The biggest problem that I'm struggling with so far is that I need the analysis to take into consideration that depending on the group, participants have studied the same words but with different methods. To give an example:

Participant #1 - "Group A" - Studied "Set Up" with the "Retrieval Method" which was "Correct"
Participant #31 - "Group B" - Studied "Set Up" with the "Conceptual Metaphor Method" which was "Incorrect"

I'm only just reading about this now and it's relatively new to me, but I believe I need to have some nested factors within the participants to focus on this difference between teaching methods?

I have a few more questions about the set up in Jamovi, but maybe I'll leave things here for the meantime. I think I need to address this nesting question first before moving on.

Thanks in advance.
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GLMM - Need some help setting up my data

Post by mcfanda@gmail.com »

Hello
you want to have a dataset in which each row represents a trial, so one row is the response of one participant to one word. Each row will have as variables the response (correct vs not correct), the method (group), the class, the participant ID (so you know to which participant the row belongs) and the word ID (so you know which word was answered)

Now you can run a generalized mixed model selecting "logistic" as type of model. Response is the dependent variable and group is the factor. Participants ID and word ID goes in "clustering variables". Then you want to put the random intercepts for participant ID and word ID.

Class would normally be another clustering variable, but because you only have four of them, it is better to put class as an additional factor.
Rageofanarchy
Posts: 2
Joined: Fri Feb 03, 2023 2:42 am

Re: GLMM - Need some help setting up my data

Post by Rageofanarchy »

Hi there,

Thanks for getting back to me. I've attached two screenshots. One is what my data setup looks like. It is exactly as you described, I believe. The first participant, participant #1, has 60 rows as they answered 20 questions on the pre, post, and delayed.

I ran the analysis, but I don't think I'm there yet. One thing you didn't mention was the three tests that they took at different stages during the study (PrePostDelayed) and the teaching method (R_CM).

As mentioned in my original post, I really need to tease out the interaction between R_CM teaching methods.

Participant #1 - "Group A" - Studied "Set Up" with the "Retrieval Method" which was "Correct"
Participant #31 - "Group B" - Studied "Set Up" with the "Conceptual Metaphor Method" which was "Incorrect"

But as the participants in each group studied the same words in different ways, I'm not sure if I need to nest this data within another factor somehow.

Thanks again.
Attachments
Screenshot 2023-02-03 at 21.34.13.png
Screenshot 2023-02-03 at 21.34.13.png (588.94 KiB) Viewed 1639 times
Screenshot 2023-02-03 at 21.30.30.png
Screenshot 2023-02-03 at 21.30.30.png (143.53 KiB) Viewed 1639 times
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GLMM - Need some help setting up my data

Post by mcfanda@gmail.com »

There is no nesting here, the two clustering variables are the ones we said. From now on, you can add all your fixed factor effects (main effects and interactions) as it was a standard ANOVA.
As far as I can tell from the screenshots, you want to include the effect of R_CM and the effect of prepostdelay, and their interaction. No problem, just add the variables as factors in your model. If the effects are within-subject, you can add these effect as random effects.
Post Reply