In most free-living bird populations moult progression and duration in individuals can not be observed fully. Instead snapshot measurements of (re)captured individuals are typically used to infer these parameters on a population level. As an additional complication, recording of moult in the field may take various forms both in terms of the subset of the population that is sampled and whether moult is recorded as a categorical state, or a (semi-)continuous progression.
moultmcmc implements a Bayesian inference framework for this class of models with the aim of (eventually) allowing the inclusion of hierarchical model structures to accommodate 1) the integration of moult data sets using different modes of recording (☑), 2) individual heterogeneity in moult timing (☑) and progression (☐), and 3) misclassified observations of non-moulting birds (☑)
moultmcmc implements fast inference for these models using Hamiltonian Monte Carlo samplers from Stan. The currently implemented models are described in detail in the vignette ‘Moult data likelihoods’.
install.packages("moultmcmc", repos = "https://pboesu.r-universe.dev")
On Mac and Windows systems this will make use of pre-compiled binaries, which means the models can be run without having to install a C++ compiler. On Linux this will install the package from a source tarball. Because of the way the Stan models are currently structured, compilation from source can be a lengthy process (15-45 minutes), depending on system setup and compiler toolchain.
moultmcmc from the github source (not generally recommended for Windows users) use the following code. This requires a working C++ compiler and a working installation of rstan:
Basic usage is described in the vignette ‘Getting started with moultmcmc’
Please note that
moultmcmc is under active development and API changes may still occur.