scalewt: Compute or scale data using (weighted) means, variances and...

In ade4: Analysis of Ecological Data: Exploratory and Euclidean Methods in Environmental Sciences

Compute or scale data using (weighted) means, variances and covariances (possibly for the levels of a factor)

Description

These utility functions compute (weighted) means, variances and covariances for dataframe partitioned by a factor. The scale transforms a numeric matrix in a centred and scaled matrix for any weighting.

Usage

covwt(x, wt, na.rm = FALSE)
varwt(x, wt, na.rm = FALSE)
scalewt(df, wt = rep(1/nrow(df), nrow(df)), center = TRUE, scale = TRUE)
meanfacwt(df, fac = NULL, wt = rep(1/nrow(df), nrow(df)), drop = FALSE)
varfacwt(df, fac = NULL, wt = rep(1/nrow(df), nrow(df)), drop = FALSE)
covfacwt(df, fac = NULL, wt = rep(1/nrow(df), nrow(df)), drop = FALSE)
scalefacwt(df, fac = NULL, wt = rep(1/nrow(df), nrow(df)), scale = TRUE, drop = FALSE)

Arguments

x	a numeric vector (varwt) or a matrix (covwt) containg the data.
na.rm	a logical value indicating whether NA values should be stripped before the computation proceeds.
df	a matrix or a dataframe containing the data.
fac	a factor partitioning the data.
wt	a numeric vector of weights.
drop	a logical value indicating whether unused levels should be kept.
scale	a logical value indicating whether data should be scaled or not.
center	a logical value indicating whether data should be centered or not.

Details

Functions returns biased estimates of variances and covariances (i.e. divided by n and not n-1)

Value

For varwt, the weighted variance. For covwt, the matrix of weighted co-variances. For scalewt, the scaled dataframe. For other function a list (if fac is not null) of dataframes with approriate values

Author(s)

Stéphane Dray [e-mail]

Examples

data(meau)
w <- rowSums(meau$spe)
varwt(meau$env, w)
varfacwt(meau$env, wt = w)
varfacwt(meau$env, wt = w, fac = meau$design$season)
covfacwt(meau$env, wt = w, fac = meau$design$season)
scalewt(meau$env, wt = w)

ade4 documentation built on Feb. 16, 2023, 7:58 p.m.