Factoring Method in EFA

Discuss statistics related things

by vincep » Sun May 07, 2017 5:53 am

jamovi is great - thanks!

A question about efa:
Which factoring method is used in the exploratory analysis?
Is it possible to specify a method?

I loaded the same data in jamovi and R, and tried all of the options in the fa function from the psych library, but can't seem to match the results in jamovi. What is jamovi doing behind the scenes?
vincep
 
Posts: 3
Joined: Sat May 06, 2017 11:37 am

by Ravi » Sun May 07, 2017 7:50 am

We use the "minimal residual" factoring method from psych::fa. Which version of psych are you using? The author of psych (William Revelle) recently fixed something in the minimal residual method (in version 1.7.5) but we are still using the previous version. The next version of jamovi will use the updated psych package.

We haven't included an option that allows you to specify a different factoring method because we want to keep the analysis as simple as possible. However, I'm happy to include it if there's an obvious need for it.
User avatar
Ravi
 
Posts: 139
Joined: Sat Jan 28, 2017 11:18 am

by bcollins » Mon Dec 04, 2017 1:14 pm

It might still be nice to allow users to select among extraction methods, especially for those transitioning from SPSS, since minimal residual is not an option in SPSS and is rarely reported in research that I can find in the social-behavioral sciences. Principal axis and maximum likelihood are far more commonly used. Not a huge deal, though, but it is one less thing for users to have to justify. This is not a principled, statisitcal critique of minimum residual extraction...it may very well be better in many cases.

It would also be handy if an option for the factor loadings table could be to sort items by factor, a la SPSS. This would make the output table more consistent with how people report results.

jamovi is fantastic. I'm commonly recommending it to individual students, and getting close to recommending it as a general alternative to SPSS for many use cases in my classes.
bcollins
 
Posts: 4
Joined: Wed Sep 06, 2017 6:09 pm

by scroon » Fri May 04, 2018 3:06 am

the one thing that would also be useful is allowing the option of seeing the variable names, in order to make sense of the factor structure.
scroon
 
Posts: 15
Joined: Tue Feb 27, 2018 3:52 am

by scroon » Fri May 04, 2018 3:17 am

Also, I have a dumbo question - in SPSS, a minimum eigenvalue is 0; in Jamovi the eigenvalues range from values BELOW 0. I can't see anything in the guide to psych:fa (http://personality-project.org/r/psych/HowTo/factor.pdf) that explains why this may be the case....
scroon
 
Posts: 15
Joined: Tue Feb 27, 2018 3:52 am

by statsnewbie » Sun Oct 21, 2018 6:54 pm

I literally made an account just to say this. I would consistently use Jamovi over SPSS for EFA if it included different factoring methods. In this case, maximum likelihood. I've never heard of minimum residuals used in EFA until now and so I'd have a harder time justifying its use to colleagues. Please add in maximum likelihood as an option :relaxed:

Jamovi is still my favorite SPSS alternative!
statsnewbie
 
Posts: 2
Joined: Sun Oct 21, 2018 6:49 pm

by Ravi » Sun Oct 21, 2018 7:19 pm

I'm curious to know what the most used factoring methods are. It's not very hard to add them, so if I know which ones are most important I'll add these.

Cheers,
Ravi
User avatar
Ravi
 
Posts: 139
Joined: Sat Jan 28, 2017 11:18 am

by Ravi » Mon Oct 22, 2018 11:34 am

Ok, I now added 'maximum likelihood' and 'principal axis factoring' as possible extraction methods (next to the 'minimum residual' method). Should be available in the next version. Is there another extraction method that you think should be added?
User avatar
Ravi
 
Posts: 139
Joined: Sat Jan 28, 2017 11:18 am

by statsnewbie » Tue Oct 23, 2018 5:43 am

Awesome! Thank you for adding those! I'm no expert on EFA so I'm not sure which ones are the most popular, but SPSS has different extraction methods in their package: PCA (but we really should never do this anyway), Unweighted least-squares method, generalized least-squares method, Alpha, and Image factoring.

A list can be found here: https://www.ibm.com/support/knowledgece ... t_ext.html

Although with Maximum Likelihood, minimum residuals, and PAF included with Jmaovi, I think it would satisfy the majority's factor extraction needs.

Thanks again!
statsnewbie
 
Posts: 2
Joined: Sun Oct 21, 2018 6:49 pm

by ftr » Wed Oct 24, 2018 9:06 pm

A quite advanced exploratory factor analysisis has been developed by LORENZO-SEVA and JOAN FERRANDO and is simply called FACTOR. http://psico.fcep.urv.es/utilitats/factor/index.html . They recommand an oblique factoring with the number of factors calculated through parallel analysis. Have a look.
ftr
 
Posts: 4
Joined: Fri Dec 08, 2017 12:29 pm

Next

Return to Statistics