mvpa2.misc.dcov.dCOV(x, y, rowvar=1, uv=False, all_est=True)

Estimate dCov measure(s) between x and y. Allows uni- or multi-variate estimations

Name dCOV was chosen to match implementation in R energy toolbox:


rowvar : int, optional

If rowvar is 1 (default), then each row represents a variable, with observations in the columns. If 0, the relationship is transposed: each column represents a variable, while the rows contain observations.

uv : bool, optional

dCov is a multivariate measure of dependence so it would produce a single estimate for two matrices NxT and MxT. With uv=True (univariate estimation) it will return estimates for every pair of variables from x and y, thus NxM matrix, somewhat similar to what numpy.corrcoef does besides not estimating within x or y

all_est : bool, True

Since majority of computation of dCor(x,y), dVar(x) and dVar(y) is spend while estimating dVar(x, y) it makes sense to estimate all of them at the same time if any of the later is necessary. So output would then consist of dCov, dCor, dVar(x), dVar(y) tuple, matching the order of energy toolbox dCOV output in R.