fcbw: Fast Computation of Pearson and Spearman Correlation...

In dsnavega/cbfsr: cbfsr: Tools for correlation-based feature selection in R

Description

Fast Computation of Pearson and Spearman Correlation Coefficient

Usage

fcbw(
  x,
  y,
  method = c("pearson", "spearman"),
  partial = T,
  weights = NULL,
  digits = 3
)

Arguments

x	a matrix or data.frame. Data is internally coerced to numeric representation.
y	vector (numeric or factor). Internally coerced to numeric representation.
method	a character stating the method to used. Default is "pearson".
partial	a logical stating if correlation among x should be account for y. Default is T.
weights	a numeric vector of weights used to computed weighted correlations.
digits	an integer defining the precision of round function. Default is 3

Details

This implementation uses matrix operations for fast computation of correlation coefficients. A drawback of this approach is that the data must be complete, that is no NA values. This function implements a fast mean imputation step where the NA values are substituted by 0 after the data been scaled as part of the algorithm for fast computation of the correlation.

Value

a list with two components, x and y, storing the inter-correlations among x and the correlations of x with y respectively. The component x is a symmetric squared matrix and the component y is a numeric vector.

Author(s)

David Senhora Navega

dsnavega/cbfsr documentation built on Jan. 1, 2021, 12:07 a.m.