R scripts accompanying the paper:
Lavoue J, Burstyn I. (2021) Evidence of Absence: Bayesian Way to Reveal True Zeros Among Occupational Exposures. Annals of Work Exposures and Health, 65(1), 84–95. https://doi.org/10.1093/annweh/wxaa086
This repository contains the code used to estimate the zero-inflated lognormal mixture model (proportion of true zeros, geometric mean, and geometric standard deviation) presented in the paper, including:
- the truncated model
- the censored Bernoulli-lognormal mixture model
- the standard censored lognormal model (baseline)
Bayesian estimation (MCMC) implemented in both JAGS and STAN.
An online application allowing reproduction of the analyses is available here: https://lavoue.shinyapps.io/Mixture_app_STAN/
Note : as of july 2026, not all calculus implemented in the Web app are available here
If you use this code, please cite the paper above.