I would like to assess the skills of a lot of persons on a few main axis. Let me explain my proposal. For each skills axes, I propose several questions. Each question gives a score from 0 to 10. Once data are collected, I perform a PCA and analyze the results. Variables space gives me the skills that may be contradictory or join. Rows space gives clusters of people, who have similar skills. What do you think of this approach? Any idea of something more academic? Thanks!
Asked
Active
Viewed 39 times
1
-
I suggest you first stick to you use standard terminology as *principal components scores* and *principal modes of variation* and then more into reexpressing them in terms of your particular problem. This will greatly simplify the interpretability of your results. Having said this, probably Factor Analysis (PCA with the addition of an explicit noise model essentially) will be more appropriate in this case. – usεr11852 Oct 29 '15 at 07:05
-
Thanks for you answer. I need to study a bit more factor analysis to understand the advantage versus PCA. About the PC score : the idea is to summarize the multiple informations of the original questions. Is this clearer? – Mic Oct 29 '15 at 07:32
-
How much of the "multiple information" is retained will dependent on your eigenvectors (your principal modes of variations) and their respective eigenvalues. The scores, the projections of your original data in the space spanned by the eigenvectors, will allow you to identify clusters. The variability of the scores will be directly related with the eigenvalues estimated. – usεr11852 Oct 29 '15 at 07:38
-
Nice! Thank a lot for this clarification. Now I need to understand more FA. Maybe I'll get back to this thread if I need more clarification. At least, you confirmed me that PCA/FA are appropriate methods to assess quantitatively people skills. Thanks a lot! – Mic Oct 29 '15 at 07:50
-
I do not think I did the confirmation part but OK! See this thread on [is there any good reason to use PCA instead of EFA?](http://stats.stackexchange.com/questions/123063/is-there-any-good-reason-to-use-pca-instead-of-efa), all answers are *really* constructive. – usεr11852 Oct 29 '15 at 08:02