Using the ggmcmc package

ggplot2, based on the grammar of graphics (Wilkinson et al. 2005), empowers R users by allowing them to flexibly crate graphics (Wickham 2009). Based on this idea, ggmcmc is aimed at bringing the design and implementation of ggplot2 to MCMC diagnostics, allowing Bayesian inference users to have better and more flexible visual diagnostic tools. ggplot2 is based on the idea that the input of any graphic is a data frame mapped to aesthetic attributes (colour, size) of geometric objects (points, lines). Therefore, in order to create a graphic the three components must be supplied: a data frame, at least one aesthetic attribute and at least one geometric object. The flexibility comes from the fact that it is very easy to extend basic graphics by including more aesthetic elements and geometric objects, and even faceting the figure to generate the same plot for different subsets of the dataset (Wickham 2009, 3). The implementation of ggmcmc follows this scheme and is based on a function (ggs()) that transforms the original input (time series of sampled values for different parameters and chains) into a data frame that is used for all the graphing functions. The plotting functions do any necessary transformation to the samples and return a ggplot object, which can be plotted directly into the working device or simply stored as an object, as any other ggplot object. Finally, getting ggplot objects as the output of the plotting functions has also a positive effect: while the defaults in ggmcmc have been carefully chosen, the user later can tweak any graph by—following the ideas of the grammar of graphics—adding other geometric figures, layers of data, contextual information (titles, axis) or applying themes. So, to sum up, the implementation is driven by the following steps: