Mixed ANOVA

Discuss statistics related things

by nwong12 » Tue Nov 27, 2018 9:45 pm

My labmate and I are trying to run a mixed ANOVA in Jamovi. We heard that we could run a GLM, but are not sure if the results we are getting are correct. We compared the results from R, using ezANOVA and got different results. Would anyone be able to advise us on how to run a mixed factorial ANOVA?

We used GAMLj and our results showed the df from the intercept, rather than each variable.

Thanks in advance!
nwong12
 
Posts: 5
Joined: Mon Nov 12, 2018 8:55 am

by mcfanda@gmail.com » Wed Nov 28, 2018 12:18 am

Hi,
first, you can see examples of mixed models applied to repeated measures designs here https://blog.jamovi.org/2018/11/13/intr ... gamlj.html, and here https://mcfanda.github.io/gamlj_docs/mi ... mple2.html.

As regards the correspondence of a mixed model results with the eZANOVA or similar packages results, it depends on the design and the way the mixed model is defined. If you post your design, we can see what is the model that better suits your needs.

As regards the df, also the df depends on how you specify the model. More specifically, the df depend on what are the random effects in the model and their variance. Thus, to understand the df, you need to clarify how the random effects are defined.
mcfanda@gmail.com
 
Posts: 144
Joined: Thu Mar 23, 2017 9:24 pm

by nwong12 » Thu Nov 29, 2018 11:19 pm

We are trying to conduct a 2 (ethnic identity) x 3 (facial feature) x 6 (emotion) mixed factorial ANOVA. The between subjects factor is ethnic identity, while the within subjects factors are facial feature and emotion. There are 54 participants in the dataset. Participants viewed a series of photos and were asked to identify which emotion was displayed, given multiple choice options. They were to rate the intensity of the emotion on a 9 point Likert scale. Our research question is: do the strengths of ethnic identity (high or low) moderate the intensity ratings of facial expressions of emotion in narrow, original, and round faces?

This is our overall design. Could you please advise? We are happy to clarify additional details regarding the study if needed. Thank you!
nwong12
 
Posts: 5
Joined: Mon Nov 12, 2018 8:55 am

by mcfanda@gmail.com » Fri Nov 30, 2018 5:38 pm

How many photos are there in each condition? When we know that, we ready to model :-)
mcfanda@gmail.com
 
Posts: 144
Joined: Thu Mar 23, 2017 9:24 pm

by nwong12 » Thu Dec 06, 2018 7:34 pm

There are 72 images in total. 36 images are of female faces, while the other 36 are of male faces. There are 12 images for each of the six emotions: anger, disgust, fear, happiness, sadness, and surprise. There are 24 images for each of the three facial types: narrow, un-manipulated, and round.
nwong12
 
Posts: 5
Joined: Mon Nov 12, 2018 8:55 am

by mcfanda@gmail.com » Mon Dec 10, 2018 11:18 am

HI
You can use GAMLj Mixed to analyze your data alright. You can also take into the account the variability due to images. There's a worked out example in the docs that explains how to analyze designs very similar to yours (https://mcfanda.github.io/gamlj_docs/mi ... mple3.html ) . You can follow the example to model the repeated measures and add your between-subjects factors as "factors" in the model.
mcfanda@gmail.com
 
Posts: 144
Joined: Thu Mar 23, 2017 9:24 pm

by nwong12 » Thu Dec 13, 2018 11:42 pm

Thank you for getting back to me. We tried analyzing with the above directions and were not able to get the appropriate ANOVA summary table. When we enter a random effect, the output is not showing up at all. Could you please advise? Thank you!
nwong12
 
Posts: 5
Joined: Mon Nov 12, 2018 8:55 am

by mcfanda@gmail.com » Fri Dec 14, 2018 2:04 pm

Can you share the omv file?
mcfanda@gmail.com
 
Posts: 144
Joined: Thu Mar 23, 2017 9:24 pm

by nwong12 » Thu Jan 31, 2019 9:01 pm

I will email the omv file to you. Thanks so much for helping us with this.
nwong12
 
Posts: 5
Joined: Mon Nov 12, 2018 8:55 am


Return to Statistics