JASP ver 0.14.1 no longer supports Chronbach's Alpha with 2 variables?
I just updated to JASP version 0.14.1 on a Mac running Catalina. Reliability analysis has now moved out of the descriptives tab and you can no longer run Chronbach's Alpha with less than 3 variables. I've done this previously, but can't anymore. I assume this is because it is based on the ANOVA which requires 3 groups. Can someone explain an alternative way to run an ICCa with 2 trials in JASP?
Comments
Hi Sodosopa,
There are multiple issues here, let's tackle them one at a time. First off: yes, reliability moved from under descriptives and is now its own module. Am I correct to think that you found the module, but you can no longer run Cronbach's alpha with fewer than 3 variables? I checked and indeed, if you put in 2 variables the output table gives an error message. This is not a bug then. Maybe there is an identifiability issue with only 2 variables -- I'll ask our expert.
Cheers,
E.J.
Hi,
this is indeed not a bug, because we prevented an analysis with fewer than 3 variables, because the if-item-dropped statistics would fail in such a case. You are right, mathematically it is possible to compute alpha with only two variable.
UnfortunateIy in order to do so in JASP 0.14.1 you would have to use the alpha formula and the correlation and do the calculation manually. Standardised alpha would then be: 2*(cor(item1, item2)) / (1+cor(item1, item2). You can obtain the correlation from the "Regression" tab in JASP.
If it is an ICC you are looking for, you could check out one of the latest nightlies (https://static.jasp-stats.org/Nightlies/). We will be offering ICCs in the next release version of JASP and the nightlies are a kind of beta-version that is constantly updated with the newest changes.
Cheers,
Julius
Hi Julius,
It seems suboptimal to prevent the entire analysis only because the "if-item-dropped" subanalysis won't work. Why not allow the standard analysis for the 2 variable case and provide an error message whenever the "if item dropped" is selected (or grey out "if item dropped" if the number of variables is 2 or less)?
E.J.
Thank you both for your responses. Yes, EJ you are correct. I also agree that this is not optimal. I don't have a problem with this, and to be honest, I would just run it in R. I do like to use JASP with my students though since it is pretty intuitive and there is no learning curve. Not to mention it's free, so they can keep using it when they are no longer affiliated with a university.
I can't ask my students to manually calculate this (or I guess I could, but it would not go well). This seems too complicated for students in what is mostly an introductory statistics course. I will likely just add a third trial to their assignment, so that it runs smoothly. This is unfortunate for them if they every need to calculate an alpha on with only 2 trials.
Thanks again for your responses.
Okay. I think this can also be part of the next release. Unfortunately, I dont think I have a better solution for you right now. Cheers, Julius