Plot posterior draws of the delay distribution

Plots the draws returned by delay_parameter_draws(), delay_summary_draws() and add_summaries(). The default, type = "parameters", draws the posterior density of each distributional parameter, and of any summary column add_summaries() added, in its own panel. type = "delay" draws the delay distribution the draws imply over a grid of delays, as the posterior median density with a ribbon between two quantiles, or as one line per draw when ndraws is given.

Usage

# S3 method for class 'epidist_delay_draws'
plot(x, ...)

# S3 method for class 'epidist_delay_draws'
autoplot(
  object,
  type = c("parameters", "delay"),
  by = NULL,
  pars = NULL,
  true_values = NULL,
  ndraws = NULL,
  probs = c(0.05, 0.95),
  max_delay = NULL,
  family = NULL,
  ...
)

Arguments

x, object

An epidist_delay_draws object, as returned by delay_parameter_draws(), delay_summary_draws() or add_summaries().

...

Passed from plot() to autoplot(). Unused otherwise.

type

Either "parameters", the default, to plot the posterior density of each parameter, or "delay" to plot the delay distribution the draws imply.

by

A character vector of columns of object that define the strata to colour by. If NULL, the default, the variables recorded on object are used. See the details.

pars

A character vector of the columns to plot when type = "parameters". If NULL, the default, the distributional parameters of the family are plotted, along with the mean, sd and quantile columns present.

true_values

A named numeric vector of true parameter values to mark with dashed vertical lines when type = "parameters". The names must be among the parameters plotted.

ndraws

The number of draws per stratum to plot the delay distribution of when type = "delay", one line each, sampled at random. If NULL, the default, the posterior median density is drawn with a ribbon between the probs quantiles instead.

probs

A numeric vector of two probabilities giving the quantiles the ribbon spans when type = "delay". Defaults to c(0.05, 0.95).

max_delay

The largest delay to evaluate the delay distribution at when type = "delay". If NULL, the default, the posterior median of the 99% quantile of the delay distribution is used.

family

A model fit with epidist(), a brms family, or the name of one, giving the delay distribution. If NULL, the default, the family recorded on object is used.

Value

A ggplot object.

Details

The strata of the draws are coloured. By default they are the unique combinations of the variables in the distributional parameter formulas, which delay_parameter_draws() records, and each .row when the draws have several rows of newdata but no such variables. Pass by to stratify by other columns of the draws. The columns in by are kept in the plot data, so the plot can be faceted by them.

The delay distribution is evaluated with the density of the family for the lognormal, gamma and Weibull families. For any other family, delays are simulated from each draw as in add_summaries() and their density is estimated with stats::density(). Either way the density is evaluated at every draw, so thin the draws with the ndraws argument of delay_parameter_draws(), or build newdata with epidist_strata(), when there are many.

plot() and autoplot() are the same function. Both need ggplot2.

The plot is drawn with ggplot2::theme_minimal() and the colour blind friendly palette the package documentation uses. Add a theme or a scale of your own to the returned plot to override either.

See also

delay_parameter_draws() and delay_summary_draws() for the draws, and plot_events() to plot the data.

Other plot: plot_events()

Examples

draws <- data.frame(
  mu = rnorm(200, 1.8, 0.05),
  sigma = exp(rnorm(200, log(0.5), 0.05))
) |>
  add_summaries(family = "lognormal", probs = 0.5)
plot(draws, true_values = c(mu = 1.8, sigma = 0.5))

plot(draws, type = "delay")

plot(draws, type = "delay", ndraws = 50)