Plot hierarchy

NOTE: This currently WIP

Function structure: function = one axis <-> one function {function1, function2} = one axis <-> select one function (function1, function2) = one axis <-> multiple functions

function_name <- function_combine[{function_select1,function_select2},function_name1,(function_same_ax1, function_same_axis2)]

ArviZ has hierarchical plot-function structure.

Level 1: Atomic axis (1 axis)

The first level functions operates against one axis with one or more variables. Each function then calls backend function to do actual plotting.

These functions include:

• plot_one_kde
• plot_one_hist
• plot_one_sample
• plot_one_box
• plot_one_rug
• plot_one_text
• plot_one_point
• plot_one_autocorr
• plot_one_rank

The common input and default values for these functions are

*arr: one or two {ndarray, str}
source: {xarray.Dataset, pandas.DataFrame,bokeh.ColumnDataSource}; optional
ax: {matplotlib.axis, bokeh.figure}; optional
labels: one or two {str, bokeh.Text}
line_kwargs: {dict}; optional
    line properties
fill_kwargs: {dict}; optional
    area properties
special_kwargs: {dict}; optional
    special plot specific properties
backend: {"matplotlib","bokeh"} 

Atomic plot returns the axis.

Level 2A: multiple parameters, one atomic plot

• plot_hpd
• plot_regression
• plot_pair
• plot_parallel
• plot_ess
• plot_density <- plot_one_kde

Level 2B: One or multiple parameters, one or multiple atomic plots

• plot_rank
• plot_ppc
• plot_joint
• plot_autocorr <- plot_grid[plot_one_autocorr]
• plot_trace <- plot_grid[plot_grid[plot_dist,{plot_one_sample,plot_one_rank}]]
• plot_forest <- plot_grid[plot_grid[{plot_box,plot_kde,plot_violin},plot_one_point,plot_one_point]]

Level X: Specialized plots (special parameters)

• plot_compare
• plot_elpd
• plot_energy <- plot_one_kde
• plot_khat
• plot_loopit
• plot_posterior <- plot_grid[(plot_dist,plot_text)]

Level Y: Dashboard / PlotGrid

• plot_dashboard
• plot_grid