JASP factor analysis: Problem with polychoric correlations
I wanted to conduct a factor analysis based on polychoric correlations with JASP 0.19.2. Variables are declared as ordinal, ranging from 0 to 2. Both principal component analysis and exploratory factor analysis do not run. I get an error message stating that estimation failed and that tetrachoric/polychoric correlations require dicotomous data. This should be true für tetrachoric correlations, but not for polychoric correlations.
Any idea how to solve this problem?
Comments
Hi,
You are right, it should in theory work for ordinal variables with three categories. Could you share your data set with me, so I can look into it? You could email it to issues@jasp-stats.org.
Julius
Thank you for the fast response. I just sent the data file to issues@jasp-stats.org.
okay, so the issue is that some variables are range restricted. These are KOM20, KOM25, and KOM31. They only have scores of 0 and 2, not 1. The underlying R package (psych) assumes that they are binary, but at the same time also assumes a binary variable has to be coded 0/1. It is a bit of a weird condition, but we did not make the package unfortunately.
Now, I assume you want those items to be treated as ordinal as well, so I have to think about a way to do that.
Thank you! I'd be happy to hear from you again if you find a solution.
Try making sure that your ordinal variables are correctly specified as "ordinal" in JASP (not just "nominal"), and double-check that your data doesn't have any missing values. Also, for polychoric correlations, ensure that you're using the correct method under the "Correlation" options.
There is a fix coming to the next release version of JASP.