Something seems wrong in Factor Module

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)

by jaume.march@udl.cat » Mon Jan 11, 2021 4:42 pm

Hi guys. Thank you very much for the work done developing jamovi.
I have a problem in the Factor module,
When running a Principal Component Analysis -based on parallel analysis, and without any options enabled in Assumption Checks, Factor Loading or Additional Output- Results show a solution (working with about 160 subjects and 36 items) implying 2-components.

At the moment everything OK!!!

But the moment I activate Bartlett's test, the results update, and a 3-component solution appears.
And, if I activate the KMO option, the 2-factor solution reappears.
If I change the activated options and leave without activating any option, the 3-factor solution remains (SIC).
And sometimes when activating component summary option, a 4-factor solution appears.

I do not understand the logic of the problem since with the same configuration of options, 3 different solutions appear in some occasions.

I would appreciate any explanation or comment.

I don't know if it matters that data are not good, it's a validation into Russian of a scale created in English (not too good), which generates a PCA or EFA models that don't meet requirements for RMSEA, CFI or TLI.

Cheers from Catalonia
jaume.march@udl.cat
 
Posts: 4
Joined: Sun Dec 27, 2020 5:52 pm

by Ravi » Tue Jan 12, 2021 12:17 pm

Hi!

So the parallel analysis:
compares the eigenvalues generated from the data matrix to the eigenvalues generated from a Monte-Carlo simulated matrix created from random data of the same size
source: https://en.wikipedia.org/wiki/Parallel_analysis

So this means it is based on simulated data, and at the moment PCA runs the analysis again once you add an option which means the data is simulated again (and thus can be different then the previous run). Because the simulated eigenvalues are different, the conclusion of the parallel analysis can change in different runs. This is usually only the case if the observed data are not very conclusive. To make sure this is what's happening you can check the scree plot option to check whether the line changes along with the conclusion.
User avatar
Ravi
 
Posts: 157
Joined: Sat Jan 28, 2017 11:18 am

by jaume.march@udl.cat » Tue Jan 12, 2021 3:23 pm

Hello and thanks for replying so fast...

Sorry, I was not at all aware of having the parallel analysis option activated, despite having written it in my post!
Most of the tests I do are CFA, and I had this scheme in mind.

I think the scree plot is exactly the same whether PCA proposes 2, 3 or 4 components / factors.
The line with yellow dots (simulations) does not vary between repetitions with different number of selected factors.

Thanks
Jaume
jaume.march@udl.cat
 
Posts: 4
Joined: Sun Dec 27, 2020 5:52 pm

by Ravi » Wed Jan 13, 2021 12:49 pm

Hi Jaume,
Is it possible to share the jamovi file with us so we can have a look what's happening? If possible, you can send it to contact [at] jamovi [dot] org
User avatar
Ravi
 
Posts: 157
Joined: Sat Jan 28, 2017 11:18 am

by Ravi » Wed Jan 13, 2021 10:23 pm

I looked at the file and I get what's happening. While the parallel analysis that decides the number of factors is rerun every time a checkbox is checked, the plot is not recreated. So the plot is based on the simulations at the time the scree plot option was checked, while the number of factors keep changing if you change add or remove a checkbox. This is of course sub optimal, so I'll try to fix this at some point.

However, it's always important to look at the scree plot to see how convincing the result is. Sometimes (as is the case with your data), it's not very clear whether there should be 2, 3, or 4 factors.
User avatar
Ravi
 
Posts: 157
Joined: Sat Jan 28, 2017 11:18 am

by jaume.march@udl.cat » Thu Jan 14, 2021 8:44 am

ok!

It's about being careful to check/uncheck the scree plot option every time

thank you very much...
Jaume
jaume.march@udl.cat
 
Posts: 4
Joined: Sun Dec 27, 2020 5:52 pm


Return to Help