1
$\begingroup$

This is sort of a repost of this question, but there weren't any answers so I will ask it again and I will try to make it clear what my problem is.

My eventual aim is to find out the relationship between one independent variable and many dependent variables. Because I have many of these dependent variables I decided to use an Exploratory Facor Analysis to first reduce the number of items to fewer latent variables. Also, all these dependent variable are questionaires investigating similar, but not the same thing. And then I wanted to check whether my independent variable can predict some of the latent variables yielded by the EFA.

Now, I wonder whether I can throw my independent variable into the EFA as well and see on which latent variable it has the highest loading on? Or is that entirely unusual to do? And would it mess up my latent variables?

$\endgroup$
  • $\begingroup$ Hi Annamarie--welcome to the site. A quick clarification question: are you more interested in using your IV to predict the the "stuff" that is similar among the questionnaires as your DV, or are you more interested in using your IV to predict the "stuff" that is unique about the similar questionnaires as your DV? $\endgroup$ – jsakaluk Aug 1 '16 at 17:48
  • $\begingroup$ Hi jsakaluk, I am not sure whether I fully understand your question. My items in the questionnaires cover different cognitive skills. First I tried an orthogonal approach with 'varimax' and if I understand it right, then this means that the factors are more unique (sharing less variance). But since different cognitive capabilities are not independent from each other, I tried the factor analysis with and oblique rotation (factors more similar?). In the end, I wanted to see, if I can predict cogniton with my IV. $\endgroup$ – Annamarie Aug 2 '16 at 13:48
  • $\begingroup$ Put another way, are you more interested in using your IV to predict--and therefore draw conclusions about--cognitive capability, in general (i.e. what these questionnaires would share in common with one another), or are you more interested in using your IV to predict--and therefore draw conclusions about--each of the idiosyncratic cognitive skills, individually? $\endgroup$ – jsakaluk Aug 2 '16 at 20:12
  • $\begingroup$ Hey jsakaluk, indeed I am more interested in specific cognitive skills. I assume that my IV will influence cognition in general, but I am trying to find out, which part of cognition is affected the most. Probably the cleanest thing to do is to extract the factors and then do the regression with the factor as DV. But I was wondering if I could also put the IV in the EFA and assess on which skills the IV has most influence judging by the loading of the IV. I am just not sure, whether that will entirely mess up the factors. $\endgroup$ – Annamarie Aug 3 '16 at 8:53
1
$\begingroup$

Well, looking at the original question it seems you are inverting dependent and independent variables. Sometimes a valid approach is to reduce independent variables dimensionality using factor analysis, and then run a regression involving your dependent variable and the projections of units over factors.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.