# What to do after running an exploratory factor analysis?

Say I asked 1000 people to evaluate 10 items about one product. The data looks like

ID item1 item2 …item10
1
2
3…..


After running an explorative factor analysis, there will be for example 3 factors created, fac_1, fac_2, fac_3.

ID item1 item2 .. item10  fac_1 fac_2 fac_3
1                         -.5     2     3
2                          1      -.8    -.2
3                         -.3     -.1    -.5


I have the impression that in the books, the capital always ends here. Yes I am clear now that there are 3 hidden structures behind my 10 items, how should I evaluate my product in this three structures (aggregate level)? How much score does my product get in this three factors?

I would like to know: How much scores my products get in this three sturctures(aggregate level), for example,

in Fac_1  score 1.2
in Fac_2  score 0.5
in Fac_3  score -0.9


How can I calculate this kind of score? Any suggestion?

What confuses me is: I have 1000*3 factor scores, what should I do next? The main question is: How does my product performs in this three factors? What should I calculate?

• Unclear what you are asking. Are you asking how to compute a score for each person for each factor? This can be a simple matter of choosing the option for saving factor scores. – rolando2 Dec 19 '14 at 2:16
• Hi @rolando2 , i updated my question, what I would like to know is the scores in a aggregate level – yue86231 Dec 19 '14 at 7:51
• Are you trying to calculate factor scores that can be used in further analysis. In SPSS, just select "Save factor scores". – Jeremy Miles Dec 19 '14 at 14:49
• If that is the case, then it is a question about software implementation and which button to push, which is off-topic here. – StasK Dec 19 '14 at 15:16
• Hi @JeremyMiles, I know how to save the factor scores, what confuses me is that all books end here. I would like to know what I could do next after having 1000*3 factor scores, after all, I can not say anything about " how does my product performs in this three structure" – yue86231 Dec 20 '14 at 21:31

Usually, after exploratory factor analysis (EFA), researchers perform confirmatory factor analysis (CFA) for validating hypothesized measurement model. And it's a good idea to do that in your case as well. However, it seems that your main question is how to estimate effect of each of your uncovered latent factors. For that, you need to perform structural model analysis, also known as path analysis. Path analysis is a major part of structural equation modeling (SEM) approach, which usually consists of EFA (if needed), CFA and path analysis. A related, but more general (umbrella) term, referring to analysis of models with latent factors, is latent variable analysis (LVM).

There are various R packages for performing SEM, including path analysis. They include sem, lavaan, OpenMx, plspm, semPLS and many others. If you would like to review more detailed information on the above-mentioned methods, please check my earlier answer on Cross Validated.

• To the question "what you should do with your scores" there is no right or wrong answer that can come from any of us. It's a matter of what you are trying to accomplish, what question(s) you want to answer, and, quite possibly, why you performed factor analysis in the first place. From the minimal information provided in the question, it seems you may at least want to characterize the central tendency (basic descriptive statistics) of each of the 3 new variables. But to leap from that basic place to trying SEM seems unjustified, and unwise. – rolando2 Dec 23 '14 at 1:42
• @rolando2: Thank you for the comment. However, I respectfully disagree with you. First, the question is "What to do after EFA". Second, I believe that it's important to provide a wider perspective (on a topic), even though it might be not represent an immediate and direct answer. Often more local issues become clearer, when they are framed in a larger unified perspective. That was the underlying idea for my answer, and, based on the OP's feedback comment, it seems that was inspirational enough and, thus, helpful enough for the OP. Therefore, I think that my answer is justified and valuable. – Aleksandr Blekh Dec 23 '14 at 2:21
• @rolando2: Sorry about a typo, past the editing time: "might be not" = "might not". – Aleksandr Blekh Dec 23 '14 at 2:28

## protected by Glen_b♦Jun 13 '17 at 7:13

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).