initializeFullZ: Construct the full Z matrix

In MarioniLab/miloR: Differential neighbourhood abundance testing on a graph

Construct the full Z matrix

Description

Using a simplified version of the n x c Z matrix, with one column per variable, construct the fully broadcast n x (c*q) binary matrix that maps each individual onto the random effect variable levels. It is not intended for this function to be called by the user directly, but it can be useful to debug mappings between random effect levels and input variables.

Usage

initializeFullZ(Z, cluster_levels, stand.cols = FALSE)

Arguments

Z	A n x c matrix containing the numeric or character levels
cluster_levels	A list that maps the column names of Z onto the individual levels
stand.cols	A logical scalar that determines if Z* should be computed which is the row-centered and scaled version of the full Z matrix

Details

To make sure that matrices commute it is necessary to construct the full n x c*q matrix. This is a binary matrix where each level of each random effect occupies a column, and the samples/observations are mapped onto the correct levels based on the input Z.

Value

matrix Fully broadcast Z matrix with one column per random effect level for all random effect variables in the model.

Author(s)

Mike Morgan & Alice Kluzer

Examples

data(sim_nbglmm)
random.levels <- list("RE1"=paste("RE1", levels(as.factor(sim_nbglmm$RE1)), sep="_"),
                      "RE2"=paste("RE2", levels(as.factor(sim_nbglmm$RE2)), sep="_"))
Z <- as.matrix(data.frame("RE1"=paste("RE1", as.numeric(sim_nbglmm$RE1), sep="_"),
                          "RE2"=paste("RE2", as.numeric(sim_nbglmm$RE2), sep="_")))
fullZ <- initializeFullZ(Z, random.levels)
dim(Z)
dim(fullZ)

MarioniLab/miloR documentation built on Oct. 18, 2024, 6:04 p.m.