compute_ev() error

[rfarouni]

Hi Natalia,

Thanks for making the code publicly available! I was trying it out on my dataset and I noticed that the function throws an error when best_numCC is smaller than the number of observations.

cca.best_numCC
Out[10]: 4

cca.compute_ev(beta_dsc)
Computing explained variance for component #1
Computing explained variance for component #2
Computing explained variance for component #3
Computing explained variance for component #4
Computing explained variance for component #5
Traceback (most recent call last):

  File "<ipython-input-9-9680903395ab>", line 1, in <module>
    cca.compute_ev(beta_dsc)

  File "rcca.py", line 53, in compute_ev
    preds, corrs = predict(data, [w[:, ccs-1:ccs] for w in self.ws], self.cutoff)

  File "rcca.py", line 172, in predict
    iws = [np.linalg.pinv(w.T, rcond = cutoff) for w in ws]

  File "/home/rick/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 1583, in pinv
    _assertNoEmpty2d(a)

  File "/home/rick/anaconda/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 222, in _assertNoEmpty2d
    raise LinAlgError("Arrays cannot be empty")

LinAlgError: Arrays cannot be empty

I was able to get it to work by replacing this line

self.numCC = nT if self.numCC is None else self.numCC

with

self.numCC = self.best_numCC

But I am not sure if that is a satisfactory fix given the rest of the code.

Best,
Rick

[nbilenko]

Hi Rick,

Thank you for using the code and for catching the bug! It is now fixed (I apologize for the delay). Your fix should be fine as well.

Cheers,
Natalia