adding discrete argument to stat functions

[behramulukir]

As stated by @santikka at #330, currently, stat functions in bayesplot, such as ppc_stat and ppd_stat, completely rely on histograms while plotting. This is problematic since histograms are not optimal for discrete data, which the user might be working with. To overcome this issue, this PR adds a discrete argument to four stat functions: ppc_stat, ppc_stat_grouped, ppd_stat, and ppd_stat_grouped. The argument is by default set to FALSE, so the functions behave as previously. However, when it is set to TRUE, it calls geom_bar and plots a bar plot.

Tasks:

• implement the argument
• update tests
• update documentation

Examples

y <- example_y_data
yrep <- example_yrep_draws
ppc_stat(y, yrep, stat = "median")

group <- example_group_data
ppc_stat_grouped(y, yrep, group, "median")

So, the cases presented in the documentation are working as they should.

Now, let's look at the example case of @santikka:

y_discrete <- rbinom(20, 1, 0.2)
yrep_discrete <- matrix(rbinom(2000, 1, prob = 0.4), 1000, 20, byrow = TRUE)
ppc_stat(y_discrete, yrep_discrete, function(x) sum(abs(diff(x))))

group_discrete <- rep(c("First Half","Second Half"), each = 10)
ppc_stat_grouped(y_discrete, yrep_discrete, group_discrete, function(x) sum(abs(diff(x))))

On the other hand, when we use the discrete argument, they are plotted nicely as bar plots.

ppc_stat(y_discrete, yrep_discrete, function(x) sum(abs(diff(x))), discrete = TRUE)
ppc_stat_grouped(y_discrete, yrep_discrete, group_discrete, function(x) sum(abs(diff(x))), discrete = TRUE)

A similar logic applies to ppd_stat versions as well.

ppd_stat(yrep, "median")
ppd_stat_grouped(yrep, group, "median")

ppd_stat(yrep_discrete, "median")
ppd_stat_grouped(yrep_discrete, group_discrete, "median")

ppd_stat(yrep_discrete, "median", discrete = TRUE)
ppd_stat_grouped(yrep_discrete, group_discrete, "median", discrete = TRUE)

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.62%. Comparing base (e7d10d9) to head (68ceb49).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #369   +/-   ##
=======================================
  Coverage   98.62%   98.62%           
=======================================
  Files          35       35           
  Lines        5729     5748   +19     
=======================================
+ Hits         5650     5669   +19     
  Misses         79       79           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jgabry]

Great, thanks @behramulukir, this looks good.

[jgabry]

Thanks @behramulukir, this is great. Merging now.