Open-source packages that turn specialist statistical methods into reproducible workflows for geometallurgical and process-data problems.
R PACKAGE / COMPOSITIONAL SIMULATION
simCoDaR
Available on GitHub · GPL-3 · early development
Generate synthetic mineral compositions or learn the multivariate structure of measured assays.
There are two ways in: start with hypothetical geological domains and mineral assemblages, then convert them into theoretical elemental and oxide assays; or fit a compositional distribution to measured assay data and simulate new samples.
The empirical workflow supports Dirichlet, logistic-normal, ILR Student-t and Gaussian-mixture models, with held-out diagnostics to compare how they reproduce the observed relationships.
The mineral-first route uses hypothetical assemblages and fixed ideal mineral chemistry. The measured-assay route fits distributions of selected compositional ratios, not absolute whole-rock grades. Neither is a calibrated deposit or an instrument, spatial-continuity or causal model. The aim is to make assumptions inspectable while testing statistical methods.
R PACKAGE / PROCESS ANALYTICS
chRonostatistics
Available on GitHub · MIT licence
Understand how process variability changes over time and with sampling interval.
Use chronological variograms, integral transforms and engineering variance-component analysis to explore ordered process data.
The package includes data checks, plotting methods and reproducible process simulations. Start with the synthetic worked example, then interpret your own results in the context of the sampling and operating process.
The process, sampling interval, time window and assumptions all matter to interpretation. These tools support analysis; they do not replace geological and operating knowledge.
