Often, when someone asks me an uncertainty visualization question I throw together an RMarkdown document to work through the problem. I have started to collect some of those examples in this repository.
If you are looking for more examples, check out the tidybayes documentation as well — the tidybayes+brms vignette in particular has some things not included here.
These are typically pretty roughly thrown together, and are in a bit of a stream-of-consciousness style. You have been warned.
-
proportions: demo of using hypothetical outcome plots (HOPs) and quantile dotplots to display uncertainty in some proportions
-
linear-regression: a variety of approaches applied to a linear regression: HOPs, density+interval, and quantile dotplots of coefficients; multiple uncertainty bands, spaghetti plots, and HOPs for uncertainty in the fit line; and posterior predictive intervals for predictive uncertainty.
-
multivariate-regression: a meandering attempt to improve on correlation heatmaps, including the use of gradients within density plots and a “dithering” approach.
-
barbarella: visualizations for a survey with a 10-point rating scale analyzed with an ordinal regression model. Includes some posterior predictive checks, HOPs (animated uncertainty), quantile dotplots, and density+interval (“half-eye”) plots.
-
arima: visualizations of forecasts for a simple autoregressive time series model. Includes spaghetti plots and HOPs, plus demos the use of not-explicitly-supported packages with
tidybayes
(in this case,bsts
—Bayesian structural time series). -
snowfall: quantile dotplots and HOPs for a snowfall prediction.
-
mtcars: Some spaghetti plots and HOPs with mtcars data.
-
mcse_dotplots: Quantile dotplots that “blur” each quantile according to its Monte Carlo standard error