Semi-Supervised Bayesian Mixture Models Incorporating Batch Correction

Semi-supervised and unsupervised Bayesian mixture models that simultaneously infer the cluster/class structure and a batch correction. Densities available are the multivariate normal and the multivariate t. The model sampler is implemented in C++. This package is aimed at analysis of low-dimensional data generated across several batches. See Coleman et al. (2022) for details of the model.

batchmix

Downloads from CRAN in the past month:

Semi-supervised and unsupervised Bayesian mixture models that simultaneously infer the cluster/class structure and a batch correction. Densities available are the multivariate normal and the multivariate t. The model sampler is implemented in C++. This package is aimed at analysis of low-dimensional data generated across several batches. See Coleman et al. (2022) for details of the model.

Advice on using the package

The main functions a user should be aware of are runMCMCChains, plotLikelihoods, plotAcceptanceRates, continueChains and processChains.

Parameters are sampled using Metropolis-Hastings so checking that the acceptance rate is important. We recommend aiming for acceptance rates between 0.1 and 0.5 for the class and batch means and batch scales ($\mu_k$, $m_b$ and $S_b$ respectively). In our testing, an acceptance rate of at least 0.4 for the class covariance matrices tended to suggest the sampler is exploring well, but smaller values were frequently associated with poor behaviour. The degrees of freedom tend to have very high acceptance rates in our testing regardless of the sampling window.

We recommend running a small number of chains for a small number of iterations to assess the acceptance rates before committing the computational resourcces to run a full analysis.

For an example of a workflow please see the short vignette.