Can't do RM ANOVA because Item "x" has no variance

Discuss statistics related things

by hcp4715 » Thu Jan 04, 2018 2:55 pm

I am trying to analyze a dataset using both JASP and jamovi. The dataset was 2 by 2 within-subject design, and the dependent variables include both reaction times and accuracy. I analyzed the reaction times and accuracy separately.

The results from both results are same for the reaction times data, but for the accuracy data, JASP has output but jamovi doesn't. The error information is "Item 'X' has no variance". I checked the data, indeed, there was one condition under which all participant get accuracy as 100%.

So, I am wondering why do they perform differently.
hcp4715
 
Posts: 5
Joined: Tue Nov 28, 2017 10:32 am

by jonathon » Thu Jan 04, 2018 10:26 pm

oh that's easy. they're not the same!

ravi wrote the RM ANOVA for JASP four (?) years ago, and then wrote the RM ANOVA for jamovi last year. they're completely separate code-bases, and don't share any code (with jamovi we came up with a much nicer framework which makes writing analyses easier, allows R syntax, etc. etc.)

in the instance you describe, throwing an error seems like the correct behaviour? (although the error message could be somewhat clearer).

cheers

jonathon
User avatar
jonathon
 
Posts: 967
Joined: Fri Jan 27, 2017 10:04 am

by hcp4715 » Fri Jan 05, 2018 12:56 pm

Thanks for your quick reply Jonathon.
Additional question: can you kindly give me some links about the no variance issue (i.e., if one condition of the data has no variance, then, performing ANOVA is not appropriate)? It seems not so intuitive to me, and directly google this error information didn't return the relevant information. Many thanks
hcp4715
 
Posts: 5
Joined: Tue Nov 28, 2017 10:32 am

by Ravi » Tue Jan 09, 2018 6:33 pm

Yeah, you are right. If one variable doesn't have any variance, the difference score still can have variance. Throwing this error message doesn't really make sense, I'll fix this. Thanks!
User avatar
Ravi
 
Posts: 141
Joined: Sat Jan 28, 2017 11:18 am


Return to Statistics