Alessandra Girlando, equimax is an orthogonal rotation. Most authoritative sources that I'm aware of recommend against using orthogonal rotations. They recommend using an oblique rotation instead.

That's because most variables are correlated to at least some extent, so using an orthogonal rotation artificially forces factors to be unrelated to each other.

As mentioned by Robert Trevethan, equamax (this is more common than equimax) is an orthogonal rotation which means that it is assumed that the constructs involved in the model under study don't have any sort of correlation, which is unrealistic.

Andy Field in his book Discovering Statistics Using SPSS (2009, 3rd edition) says "In practice, there are strong grounds to believe that orthogonal rotations are a complete nonsense for naturalistic data, and certainly for any data involving humans (can you think of any psychological construct that is not in any way

correlated with some other psychological construct?) As such, some argue that orthogonal rotations should never be used."

On the other hand, direct oblimin is an oblique rotation with some sort of flexibility (if you have some information about construct's correlation you can do adjustments using a constant called delta in the analysis's process). As a default, this constant is 0 (zero).

Another advantage is that the final part of the oblimin analysis' output is a correlation matrix between the factors. This matrix contains the correlation coefficients between factors, giving interesting information about construct's relations.

I usually apply direct oblimin rotation my tests.

In time, I have a pdf version of Andy's book. If you want, I can send you by here.

Good luck with your researches.

I would only like to add that you can think of orthogonality as a special case of an oblique rotation, if the factors are uncorrelated. So, with an oblique rotation you will see/detect both, correlated and uncorrelated cases. Should it be the case that the factors are really uncorrelated (which will presumably not happen in real world data) then you have nothing lost with a oblique rotation (for simplicity you could add a othogonal factor analysis the0, but which is not necessary), but you will miss the information about possible dependencies with an orthogonal FA.