Plotting estimates from multiple models

[eteitelbaum]

Hello,

I am interested in plotting posterior parameter estimates from multiple models in the same graph as this person describes here. Is there an easy way to do this?

Thanks!

[jgabry]

Good question. We don't have a built-in way to do this at the moment, but it seems like a feature we should consider adding.

Until then here's one way to do it using bayesplot's mcmc_intervals_data() function (which gives you the data bayesplot computes to pass to ggplot) in combination with a few ggplot functions:

# simulate having posteriors for two different models each with parameters beta[1],..., beta[4]
posterior_1 <- matrix(rnorm(4000), 1000, 4)
posterior_2 <- matrix(rnorm(4000), 1000, 4)
colnames(posterior_1) <- colnames(posterior_2) <- paste0("beta[", 1:4, "]")

# use bayesplot::mcmc_intervals_data() function to get intervals data in format easy to pass to ggplot
library(bayesplot)
combined <- rbind(mcmc_intervals_data(posterior_1), mcmc_intervals_data(posterior_2))
combined$model <- rep(c("Model 1", "Model 2"), each = ncol(posterior_1))

# make the plot using ggplot 
library(ggplot2)
theme_set(bayesplot::theme_default())
pos <- position_nudge(y = ifelse(combined$model == "Model 2", 0, 0.1))
ggplot(combined, aes(x = m, y = parameter, color = model)) + 
  geom_point(position = pos) +
  geom_linerange(aes(xmin = ll, xmax = hh), position = pos)

That makes a plot that looks like this:

[eteitelbaum]

Yes, that works. Thanks for your quick response!

[jgabry]

Ok great glad that works! I'll leave this open since this seems like a useful feature to add to the package instead of requiring this workaround.

[aaldoh]

Hi I was wondering if there was a way to achieve a similar plot but for density plots? I'm not entirely sure how to use the data from mcmc_areas_data in ggplot to use a similar workaround.

[gilleslijnzaad]

I have the same question as @aaldoh: I would like to plot multiple density estimates (e.g. for a prior and a posterior) in the same graph. Is this possible in any way using bayesplot?