dot-cca: Canonical Correlations

In giac01/ccatools: Tools for Canonical Correlation Analysis

Description

Sightly quicker than stats::cancor, and allows you to easily fit cca models in one dataset and find predicted variates/canonical correlations in another dataset.

The number of predictor (X) and outcome (Y) variables are denoted by P1 and P2 below, and N is sample size.

The function also allows the user to rotate the raw coefficients using Procrustes Analyses to target matrices (ProcrustX & ProcrustY), prior to estimating variates.

Usage

.cca(
  X_FIT,
  Y_FIT,
  X_PRED = NULL,
  Y_PRED = NULL,
  ncomp = 10,
  ProcrustX = NULL,
  ProcrustY = NULL,
  SafetyChecks = TRUE
)

Arguments

X_FIT	Numeric Matrix [N, P1] containing the training dataset predictor variables.
Y_FIT	Numeric Matrix [N, P2] containing the training dataset outcome variables.
X_PRED	Numeric Matrix [N, P1] containing the testing dataset predictor variables. Variables should be ordered in the same way as for X_FIT.
Y_PRED	Numeric Matrix [N, P2] containing the testing dataset outcome variables. Variables should be ordered in the same way as for Y_FIT.
ncomp	Numeric Scalar. Number of CCA components to keep in analyses. Must be equal to or less than min(P1,P2).
ProcrustX	Numeric Matrix [ncomp, P1] containing target matrix for Procrustes Analysis. Will align raw coefficient matrix to ProcrustX target matrix.
ProcrustY	Numeric Matrix [ncomp, P2] containing target matrix for Procrustes Analysis. Will align raw coefficient matrix to ProcrustY target matrix.
SafetyChecks	Checks the input provided for mistakes (default = FALSE).

Value

A list containing the following components

• xcoef - Estimated raw coefficients (CCA weights) for the x (predictor) variables.

• ycoef - Estimated raw coefficients (CCA weights) for the y (outcome) variables.

• variates - Variates (latent variable scores), estimated from the raw coefficient in X_FIT/Y_FIT. cbind.data.frame(xvariates, yvariates). If X_PRED and Y_PRED are provided, then variates (and xvariates/yvariates) will be the predicted latent variable scores from X_PRED/Y_PRED matrices.

• xvariates - Variates (latent variable scores) for predictor variables.

• yvariates - Variates (latent variable scores) for outcome variables.

• cc_pred - Predicted Canonical Correlations, estimated using from X_PRED & Y_PRED if provided. If no PRED matrices are specified, then this simply returns the estimated canonical correlations from the training datsets (X_FIT/Y_FIT).

• cc_fit - Estimated Canonical Correlation estimated from the training datasets (X_FIT/Y_FIT).

giac01/ccatools documentation built on July 15, 2021, 4:33 a.m.