erfe (version 0.0.1)

dexpectilizeMatR: Dexpectilize a matrix according the a single asymmetric point \(\tau \in (0, 1)\).

Description

This function is part of the erfe package. It de-expectilizes a matrix of data vertor-wise, which means subtracting the expectile of level \(\tau \in (0, 1)\) to every vector of the matrix column-wise. When \(\tau=0.5\) then the process of de-expectilizing corresponds to the process of deamining the matrix column-wise. That is, subtracting the mean-column from the column vector.

Usage

dexpectilizeMatR(ymat, aweight, panSizeVec)

Value

Return a de-expectilized matrix of the matrix ymat.

Arguments

ymat

Numeric matrix to de-expectilize column-wise.

aweight

Numeric vector of individual asymmetric weight.

panSizeVec

Numeric vector of individual panel size.

Author

Amadou Barry,

References

Barry, Amadou, Oualkacha, Karim, and Charpentier Arthur. (2022). Weighted asymmetric least squares regression with fixed-effects. arXiv preprint arXiv:2108.04737

Examples

set.seed(13)
temps_obs <- 5
n_subj <- 50
id <- rep(1:n_subj, each=temps_obs) 
asym <- 0.5
panSizeVec <- unname(unlist(lapply(split(id, id), function(x) length(x)))) 
ymat <- matrix(NA, nrow = n_subj * temps_obs, ncol = 5)
ymat <- matrix(mvtnorm::rmvnorm(n_subj * ncol(ymat),
sigma = diag(rep(1, temps_obs))), ncol = ncol(ymat))
aweight <- rep(asym, temps_obs * n_subj)
aweight[!(ymat[, 1] > mean(ymat[, 1]))] = 1 - asym
dexpectilizeMatR(ymat, aweight, panSizeVec)