search

gallantlab / pyrcca

Regularized kernel canonical correlation analysis in Python
Other
240 stars 72 forks source link

decreasing order of the correlation coefficients #4

Open HoltzO opened 6 years ago

HoltzO commented 6 years ago

Hi,

Thank you so much for sharing the Pyrcca software. this is exactly what I was looking for.

I have some questions (not sure this is the right place but I'll try anyway):

  1. I am missing statistics of the estimated model. other then in the compute_ev, are there any other statistics measures that are computed? such as F-p-value or wilks lamda?
  2. for some reason when using kernel and regularization, I don't necessarily get a decreasing order of the correlation coefficients. any idea why?

thank you very much,

Hadas

nbilenko commented 6 years ago

Hi Hadas,

  1. We do not provide additional statistical metrics at this time. I would recommend using scikit-learn for computing additional metrics.

  2. Are you referring to correlations between the datasets? The order of the components is determined using decreasing eigenvalues. The correlations between dataset projections may not be strictly decreasing in the kernel case with regularization.