binarize_community_data()

binarize_community_data

R Documentation

Binarize Community Data Matrix

Description

Converts a numeric community matrix of proportions (or counts) to a binary presence-absence matrix. Any value strictly greater than zero is recoded as ‘1L’; zeros remain ‘0L’. The resulting integer matrix is suitable for ‘filter_constant_taxa()’ when the model uses a binomial error family.

Usage

binarize_community_data(data_community_matrix = NULL)

Arguments

data_community_matrix

A numeric matrix with samples as rows and taxa as columns, as returned by ‘prepare_community_for_fit()’. Values must be non-negative.

Details

Pre-binarization before ‘filter_constant_taxa()’ is essential when using a binomial error family: a taxon recorded at non-zero varying proportions in every sample has positive SD on the proportion scale but becomes a constant-1 column after binarization inside the model, causing implicit intercept saturation. Applying this function first ensures that ‘filter_constant_taxa()’ removes such taxa before they reach the model.

For other error families (e.g., Gaussian, future hurdle models) the raw proportional matrix should be passed to ‘filter_constant_taxa()’ directly; use the ‘error_family’ configuration key to control this choice in the pipeline.

Value

An integer matrix of the same dimensions and dimnames as the input, with all non-zero values replaced by ‘1L’.