statsmodels.multivariate.factor_rotation.promax

statsmodels.multivariate.factor_rotation.promax(A, k=2)[source]

Performs promax rotation of the matrix \(A\).

This method was not very clear to me from the literature, this implementation is as I understand it should work.

Promax rotation is performed in the following steps:

  • Deterine varimax rotated patterns \(V\).

  • Construct a rotation target matrix \(|V_{ij}|^k/V_{ij}\)

  • Perform procrustes rotation towards the target to obtain T

  • Determine the patterns

First, varimax rotation a target matrix \(H\) is determined with orthogonal varimax rotation. Then, oblique target rotation is performed towards the target.

Parameters
Anumpy matrix

non rotated factors

kfloat

parameter, should be positive

References

[1] Browne (2001) - An overview of analytic rotation in exploratory factor analysis

[2] Navarra, Simoncini (2010) - A guide to emprirical orthogonal functions for climate data analysis