ChainsMakie.jl

Appearance

forestplot ​

A forestplot of Chains displays point summaries of all samples per parameter with horizontal lines representing one or more credible intervals.

Forest plots allow for easy comparison of posterior distributions across multiple parameters. This, forest plots are better suited for reporting and interpretation than diagnosing issues in MCMC sampling because they collapse samples across chains.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig, ax, plt = forestplot(chains)
fig

It is possible to plot a subset of the parameters by passing their names as the second argument:

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 3, 4), [:A, :B, :C])
forestplot(chains, [:A, :B])
(Scene(2 children, 0 plots), Axis (1 plots), Plot{ChainsMakie.forestplot, Tuple{Vector{Base.ReshapedArray{Float64, 1, AxisArrays.AxisMatrix{Float64, Matrix{Float64}, Tuple{AxisArrays.Axis{:iter, StepRange{Int64, Int64}}, AxisArrays.Axis{:chain, UnitRange{Int64}}}}, Tuple{}}}}})

Attributes ​

ci ​

You can choose which credible intervals should be plotted by passing a custom vector of values between 0 and 1. The proportion outside the ci will be distributed equally on the lower and upper tails of the summarised distribution.

Defaults to [0.95, 0.90].

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig, ax, plt = forestplot(chains; ci = [0.99, 0.95, 0.89])
fig

point_summary ​

Controls which function will be used to calculate the point_summary. Any function can be used so long as it returns a single Real number when applied to a vector Reals.

Defaults to median.

julia
using ChainsMakie, CairoMakie
import StatsBase: mean
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig, ax, plt = forestplot(chains; point_summary = mean)
fig

colormap ​

Specifies the colormap to be used for coloring the different quantiles.

Defaults to :viridis.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig, ax, plt = forestplot(chains; colormap = :plasma)
fig

min_width ​

Controls the linewidth of the narrowest interval.

Defaults to 4.

max_width ​

Controls the linewidth of the widest interval.

Defaults to 8.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 5, 4), [:A, :B, :C, :D, :E])
ci = [0.99, 0.95, 0.89, 0.8]
fig, ax, plt = forestplot(chains; ci, min_width = 3, max_width = 12)
fig