gemmr.estimators.SVDCCA

class gemmr.estimators.SVDCCA(n_components=1, covariance='empirical', scale=False, std_ddof=1, cov_out_of_bounds='nan', normalize_weights=True)

Canonical Correlation Analysis estimator based on singular value decomposition.

Parameters:
  • n_components (int >= 1) – number of between-set components to estimate
  • covariance (str) – must be ‘empirical’
  • scale (bool) – whether to divide each feature by its standard deviation before fitting
  • std_ddof (int >= 0) – when calculating standard deviations and covariances, they are normalized by 1 / n-std_ddof
  • cov_out_of_bounds (str) – if fitting results in a canonical correlation > 1, which indicates some problem, potentially that too few samples were used raise an error if cov_out_of_bounds=='raise', set association strengths and weight vectors to np.nan if cov_out_of_bounds=='nan', or ignore the problem if cov_out_of_bounds == 'ignore'
  • normalize_weights (bool (default True)) – If normalize_weights == False weights are calculated as in Härdle and Simar (2015). In this case they are not normalized (i.e. || w ||_2 != 1). Set normalize_weights to True to get normalized weights.
corrs_

contains the canonical correlations. This is the quantity that’s maximized by CCA

Type:np.ndarray (n_components,)
assocs_

Identical to corrs_. assocs_` is the common identifier used in in SVDPLS, SVDCCA, NIPALSPLS and NIPALSCCA for the association strength that is optimized by each particular method

Type:np.ndarray (n_components,)

References

Härdle and Simar, Applied Multivariate Statistical Analysis, Chapter 16, Springer (2015)

__init__(n_components=1, covariance='empirical', scale=False, std_ddof=1, cov_out_of_bounds='nan', normalize_weights=True)

Initialize self. See help(type(self)) for accurate signature.

Methods

__init__([n_components, covariance, scale, …]) Initialize self.
fit(X, Y[, copy]) Fit the estimator.
fit_transform(X, Y, **fit_params) Fit the estimator and return the resulting scores
get_params([deep]) Get parameters for this estimator.
score(X, Y[, ftr]) Returns the pearson correlation of the ftr-th canonical variates (scores).
set_params(**params) Set the parameters of this estimator.
transform(X, Y[, copy]) Apply the previously fitted estimator to new data.