This video is the first of a five-part series on using the mvgam R package for Bayesian inference of dynamic time series models. In this video, we introduce the package mvgam by using a range of dynamic State-Space models to analyse a time series of disease counts. This video covers the basics of mvgam, including:
1. Data preparation and visualisation
2. Inspection of default prior distributions
3. Updating priors and fitting models using Stan as the backend
4. Conditional effect plots
5. Residual analysis
6. Prediction and forecasting
About the mvgam package:
The goal of mvgam is to fit Bayesian (Dynamic) Generalized Additive Models, using the probabilistic programming language Stan for MCMC inference. This package constructs State-Space models that can include highly flexible nonlinear predictor effects for both process and observation components by leveraging functionalities from the impressive brms and mgcv R packages. This allows mvgam to fit a wide range of models, including hierarchical ecological models such as N-mixture or Joint Species Distribution models, as well as univariate and multivariate time series models with imperfect detection.
Where to find out more information:
- The R code used for this video: gist.github.com/nicholasjclark/c8f2fa827da233c810d1b2103fdcf7b7
- The mvgam package website: nicholasjclark.github.io/mvgam/
- A blog post on fitting Gaussian Processes with autocorrelation: ecogambler.netlify.app/blog/autocorrelated-gams/
- A blog post on models with time-varying seasonality: ecogambler.netlify.app/blog/time-varying-seasonality/
- The Stan Discourse: discourse.mc-stan.org/