Small bug (or strange behavior) in ANOVA/ANCOVA

Discuss the jamovi platform, possible improvements, etc.
Post Reply
User avatar
mcfanda@gmail.com
Posts: 452
Joined: Thu Mar 23, 2017 9:24 pm

Small bug (or strange behavior) in ANOVA/ANCOVA

Post by mcfanda@gmail.com »

Hi
I notice a small bug in ANOVA/ANCOVA modules, (it might be intended). In Assumptions check there is the Levene test that test the equivalence of variances across groups. At the moment, it checks all groups defined by the combinations of levels in the variables put in factors. However, if one changes the model in "Model" and remove one of the factor from the model (while keeping the variable in the "factors" field), the Levene's test still consider the factor levels in the test. I guess it would be better if it tested only the combinations of groups in the model.

cheers
User avatar
Ravi
Posts: 194
Joined: Sat Jan 28, 2017 11:18 am

Re: Small bug (or strange behavior) in ANOVA/ANCOVA

Post by Ravi »

Hi Marcello,

Yeah, I guess technically this behavior is a bit weird and it should follow the model builder. What's even weirder though is why someone would add a variable to the factors and then remove it again in the model terms :stuck_out_tongue:. Could you make an issue on our github page (https://github.com/jamovi/jamovi/issues)?

Cheers
User avatar
mcfanda@gmail.com
Posts: 452
Joined: Thu Mar 23, 2017 9:24 pm

Re: Small bug (or strange behavior) in ANOVA/ANCOVA

Post by mcfanda@gmail.com »

HI
you're right, I can think of one case which might make sense here: when there are missing values. ifl you select your predictors in the variables panel and then run different analyses with different models, you're sure that all the models are estimated on the same sample. I', going to open a github issue.
thanks
Post Reply