Hello,

Factor analysis are generally used to determine unobserved variables from observed variables. This types of analyses can be used to ensure of psychometric qualities of a scale.

I think that to estimate the factors leading students to pursue their studies, there are two main solution :

* Regression Analysis

* Structural Equation Modeling (SEM)

These both of these types of analyses can provide the influence of variables on another variable. But SEM offers an unique framework for analyziyng measurement model and structural model.

Hello Blanca,

As Nicholas indicates, it would more sense to try a different analytic method.  If your key question of interest is, what variables help determine why someone chooses Mexico vs. anyplace else as a destination, then any of these methods could be informative:

Discriminant analysis (DV = Mexico vs. other destination; IVs = your push/pull measures)

Logistic regression (DV = Mexico vs. other; IVs = push/pull measures)

SEM (whether with or without latent variables): Which push/pull measures help explain differences in destination status (Mexico vs others)? 

Good luck with your work.

If your core goal is to create a measure of the concept that interests you, then Factor Analysis will tell you whether a set of items all measure the same thing or whether they "load" on multiple underlying factors.

Alternatively, if your goal is to use the measure that you produce as either an independent or dependent variable, then regression or SEM would be appropriate as a next step.

All of the above answers offer further insight into your issue.

Factor analysis is a method, as David said, to see if measures of your concept are related to the underlying factor you expect. It is very useful in diagnosis/testing of a measure you might have in mind.  This is useful if the evaluative measures (items?) have not been tested prior and found to be robust in measuring the determining factors of interest.

Regression techniques give you the opportunity to develop a predictive model, and one of its great power is in the understanding of the variance randomness of the data and any non linearity that might exist in the relationship.

Structural Equation Modelling, does allow you to see the strength of relationships between constructs. between constructs and your measurement items, and to develop at descriptive model of the environmental factors driving the process.

However these techniques intersect. 

if your items (measures) don't load strongly, in factor analysis, on one vs another underlying factor then your measure is not measuring a single underlying influencing factor (issue determinant etc).  The result will be weak /soft loading in structural equation modelling at least in many cases, 

If the relationship is non linear then a similar issue can occur in the SEM model in that loading might not give strong relationships, or at least you will learn nothing about non linearity ie it will not be captured in a visual way, perhaps misleading you.

I am sure others on here will be more knowledgeable than me on these issues

What I feel is useful, is to use a tested or validated set of measurement items if they are available and if they are not then perhaps, see that a multiple approach is needed to get insight into the issues.

Structural Equation Modelling would be one of your best approaches since you seem to have a set of potential factors.  Could use Confirmatory Factor Analysis (rather than exploratory FA) but that seems like overkill.