Variables and Factor Analysis 2012

This is an updated version of the old "Variables and Factor Analysis" video. The old video had a couple slides out of place.

[drranaayyub]

Dear Dr. James,
Do i need to include all the moderators in EFA or do i just need to include independent and dependent variables only? Someone told me that we dont need to include moderators for your help inadvance.

[faribapakravesh]

It was great.thank you James.
How can we do Exploratory Factor Analysis when we have more than 50 variables?

[Gaskination]

Latent factors? no. You need an absolute minimum of 2 items to comprise a latent factor. So, that would mean at least 14 if there are seven factors. However, they could have two latent factors of four and three items each and then five individual (observed) variables.

[drranaayyub]

What a great Thanks for this effort. great

[Gaskination]

Correct, you cannot get model fit in SPSS. So, you could not complete the entire CFA in SPSS, only the validity and reliability parts.
You are correct on your fixing the number of factors comment.

[Gaskination]

No. Amos does not handle 2nd order formative variables (or first order ones) very well. I have seen it done, but it comes with all sorts of caveats and limitations. I would recommend PLS for formative models. I have a video that shows how to handle 2nd order formative variables in SmartPLS.

[Gaskination]

You can exclude categorical moderators that you will be using as multi-group moderators. You may optionally include the continuous moderators that you will use as interactions. You do this in order to demonstrate that the moderating variable is sufficiently reliable and distinct from the independent variables it is interacting with.

[erikpaolo]

Hello! Just one clarification regarding factor analysis. So, bottom line is that, you cannot do confirmatory factor analysis (CFA) in SPSS, right (you have to use AMOS for CFA)? If so, then what is the point of declaring the number of factors during factor analysis in SPSS (in which SPSS has that option of "fixed number of factors")? Thanks in advance!

[Gaskination]

You can actually get most of what you need in SPSS using the pattern matrix, factor correlation matrix, and the cronbach's alphas. So, you could do a CFA in SPSS. Declaring the number of factors in SPSS is useful if you are not sure where the eignenvalues are going to come in. SPSS defaults to eigenvalues > 1, but you might have factors with eigenvalues just below that threshold. So, fixing them to a certain number is useful if you already know how many you expect to find.

[erikpaolo]

Oh ok. My professor just asked me about it since he said I cannot do CFA in SPSS because SPSS cannot report the usual model-fit numbers similar to that in AMOS. So in his mind, it's not CFA.
"So, fixing them to a certain number is useful if you already know how many you expect to find." --> especially if there is already a theoretical basis to say that for example, Construct A consists of three factors, and each factor say has four items. Correct?

[LindaPlotnickPythonVideos]

Can one do a CFA in AMOS with a 2nd order formative construct? Thanks in advance for your answer.

[Gaskination]

Yikes! You will need a large sample size. Or, you can do a few constructs at a time (assuming you already have constructs designed for each set of items). Then once you weed through those constructs, do another set, then another set, then put all the remaining items together for a final EFA. I have had to do this on occasion. I once had over 80 items...

[tajamulislam]

Before doing EFA or CFA, is it necessary to reverse score the negatively phrased items?

[mahasarwar]

What should I do if my model has a combination of reflective and formative variables?
Few variables are fully reflective, few are a combination of formative and reflective

[Aminah]

Is it credible for an author to have 7 factors from a 12 item survey?