ChainsMakie.jl

Appearance

autocorplot ​

The autocorplot shows the autocorrelation of MCMC samples at different lags.

The autocorrelation of MCMC samples contains information about the efficiency of the sampling algorithm. Uncorrelated samples at short lags indicate efficient sampling, while high autocorrelations typically indicate problems in the exploration of the target distribution.

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

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])
autocorplot(chains, [:A, :B])

Attributes ​

lags ​

Controls at which lags the autocorrelation will be calculated.

Defaults to 0:20.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig = autocorplot(chains; lags = 0:5:100)
fig

color ​

Controls which colors will be used to color the samples from each chain.

Defaults to Makie's wong_colors palette and automatically switches to colormap = :viridis for more than seven chains.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 4), [:A, :B])
fig = autocorplot(chains; color = first(Makie.to_colormap(:tab20), 4))
fig

colormap ​

Controls which colormap will be used to color the samples from each chain.

Defaults to the :viridis palette.

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

linewidth ​

Controls the linewidth of the autocorrelation plot of each chain.

Defaults to 1.5.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 6), [:A, :B])
fig = autocorplot(chains; linewidth = 1.0)
fig

alpha ​

Controls the opacity of the autocorrelation plot of each chain.

Defaults to 1.0.

julia
using ChainsMakie, CairoMakie
import MCMCChains: Chains
chains = Chains(randn(300, 2, 6), [:A, :B])
fig = autocorplot(chains; alpha = 0.8)
fig