Oblimin or varimax (factor analysis) ? How to choose ?

[Bruno Daucé]

Hi !

Is there one way to choose between varimax or oblimin rotation ?

I have heard that we have to choose oblimin when correlation between
factors was more than 0,15. But i do know why.

Is there anybody whos has a suggestion ?

Bruno

[Richard F Ulrich]

Bruno Daucé (bruno...@iname.com) wrote:

: Is there one way to choose between varimax or oblimin rotation ?

: I have heard that we have to choose oblimin when correlation between
: factors was more than 0,15. But i do know why.

: Is there anybody whos has a suggestion ?

The question is not quite right.

Factors are extracted as uncorrelated. Varimax is a rotation that
keeps them uncorrelated. If the "true correlation" that you imagine
to exist is small (less than .15, for instance), then the idea is:
there will be no practical difference between the correlated solution
and the solution that you will see with varimax.

Oblimin shows you a solution that allows the axes to be correlated.
Thus, it can give better-looking loadings. The "better-looking"
solutions that I have seen, in other people's examples, have used
variables with many intercorrelations greater than .70, and have
allowed only a small amount of factor-correlation in the results.

--
Rich Ulrich, biostatistician wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html Univ. of Pittsburgh