[Bug]: EFA and PCA issues with ordinal and binary data #3163
Description
JASP Version
0.19.2
Commit ID
jasp-stats/jasp-desktop@410d926
JASP Module
Factor
What analysis are you seeing the problem on?
EFA, PCA
What OS are you seeing the problem on?
macOS Silicon
Bug Description
The following happened (from the forum https://forum.cogsci.nl/discussion/comment/30883#Comment_30883):
Trying to do an EFA with the dataset that is ordinal with three response categories (0, 1, 2) and choosing poly/tetrachoric correlations. However, some of the variables had restricted range, that is, responses were only 0 and 2. The way we do the poly/tetra correlations is to use mixedCor from the psych package which detects the scaling of the involved variables and automatically calculates the correct correlation. So for the restricted variables it assumed they were binary. Unfortunately the package also seems to assume a binary variable has to be 0/1, otherwise it fails.
Expected Behaviour
not fail and treat the variables as they are scaled in JASP, so ordinal should stay ordinal even if there are only two response categories.
Steps to Reproduce
see above.
Log (if any)
No response
More Debug Information
Now, we would not want to fix the R package, but actually treat each variable as it is treated in JASP. So if the variable is ordinal, we can specify that for the psych function and not have it run its automatic detection. We need to be careful to transform the variables to numeric though.
Final Checklist
- I have included a screenshot showcasing the issue, if possible.
- I have included a JASP file (zipped) or data file that causes the crash/bug, if applicable.
- I have accurately described the bug, and steps to reproduce it.