Issue with EFA explanatory parallel oblimin Factoranalysis

[dagiovanni]

Template for bug reports

• JASP version: 8.5.1
• OS name and version: Microsoft Windows 7
• Analysis: explanatory parallel oblimin Faktoranalysis
• Steps to reproduce: i did an EFA in JASP 8.4. (Issue with EFA explanatory parallel oblimin Faktoranalysis) and dicovered other plots and Statistiks in the new Version 8.5.1. I used the same data and same options in jasp 8.4 and in Jasp 8.5.1. The Scree plot looks different and other values of the chi-square are discoverd. In my opinion the line of the simulated date is wrong in comparision to the Jasp 8.4 Scree plot. but iam not sure. Which analysis can i trust?

Template for enhancement requests

• Enhancement:
• Purpose:

[boutinb]

@dagiovanni Can you give us the Jasp & Data files you use (in ZIP format)?

[dagiovanni]

Hi,

sure , can i send it? Which email or upload?

[boutinb]

Hi @dagiovanni I got the file (it is better to use GitHub for this, just by attaching the file on a comment, like this:
debugfile.zip

I can reproduce your problem. The researchers who can answer you, will be back next week. I will ask them.

[vandenman]

Hi @dagiovanni, I've taken a look at the issue and it seems like we've unintentionally changed the behavior for missing values of both EFA and PCA from JASP 0.8.4 to JASP 0.8.5. In JASP 0.8.4 pairwise deletion was done whereas in JASP 0.8.5 listwise deletion is done.

This will be changed in the next release where we'll create a separate option for listwise/ pairwise deletion.

For comparison, the following R code should reproduce the output from both JASP versions:

# install.packages("psych")
dat <- haven::read_sav("~/debugfile.sav")
result <- psych::fa.parallel(dat) # detect number of factors
nFactor <- result$nfact # should be 4
# JASP 0.8.4
Results <- psych::fa(dat, nFactor, rotate = "promax")
Results$STATISTIC
# [1] 1048.168

# JASP 0.8.5.1
Results <- psych::fa(dat[complete.cases(dat), ], nFactor, rotate = "promax")
Results$STATISTIC
# [1] 975.0908