Analytic non-establishment class likelihood for COVID5001-style analyses

Computes the class-conditional likelihood for a non-establishment or nonproductive class whose expected viral load is a fixed fraction of the observation-specific LOD. In the COVID5001 manuscript analysis, the default mean was LOD / 2 on the natural viral-load scale.

Usage

mbmm_noise_floor_loglik(obs, sigma, lod_fraction = 0.5, log_scale = FALSE)

Arguments

obs

Subject observations with columns y, lod, and optionally target and censored.

sigma

Measurement-error SD on the log10 likelihood scale. Either a scalar or a target-named numeric vector.

lod_fraction

Expected non-establishment mean as a fraction of the LOD. The manuscript default is 0.5.

log_scale

Whether obs$y and obs$lod are already on the log10 scale. When FALSE, the non-establishment mean is constructed as lod_fraction * lod on the natural scale and then log10-transformed.

Value

A scalar log-likelihood under the analytic non-establishment model.

Details

This class is analytic and does not require a reference trajectory bank.