impRZalr: alr EM-based Imputation for Rounded Zeros

Description

A modified EM alr-algorithm for replacing rounded zeros in compositional data sets.

Usage

impRZalr(x, pos = ncol(x), dl = rep(0.05, ncol(x) - 1), eps = 1e-04, maxit = 50, bruteforce = FALSE, method = "lm", step = FALSE, nComp = "boot", R = 10, verbose = FALSE)

Arguments

Compositional data

pos

Position of the rationing variable for alr transformation

Detection limit for each part

eps

convergence criteria

maxit

maximum number of iterations

bruteforce

if TRUE, imputations over dl are set to dl. If FALSE, truncated (Tobit) regression is applied.

method

either “lm” (default) or “MM”

step

if TRUE, a stepwise (AIC) procedure is applied when fitting models

nComp

if determined, it fixes the number of pls components. If “boot”, the number of pls components are estimated using a bootstraped cross validation approach.

number of bootstrap samples for the determination of pls components. Only important for method “pls”.

verbose

additional print output during calculations.

Value

xOrig: Original data frame or matrix
xImp: Imputed data
wind: Index of the missing values in the data
iter: Number of iterations
eps: eps

Details

Statistical analysis of compositional data including zeros runs into problems, because log-ratios cannot be applied. Usually, rounded zeros are considerer as missing not at random missing values. The algorithm first applies an additive log-ratio transformation to the compositions. Then the rounded zeros are imputed using a modified EM algorithm.

Examples

Run this code


data(arcticLake)
x <- arcticLake
## generate rounded zeros artificially:
x[x[,1] < 5, 1] <- 0
x[x[,2] < 47, 2] <- 0
xia <- impRZalr(x, pos=3, dl=c(5,47), eps=0.05)
xia$xImp