88 create a rejection sampling function to sample from non standard distributions

[kgoldfeld]

I've created genDataDensity and addDataDensity so that it is now possible to generate data from an arbitrary distribution specified by a vector of integers:

base_data <- c(1, 2, 2, 2, 2, 2, 3, 3, 4, 4, 4, 5, 6, 6, 7, 7, 7, 8, 9, 10, 10, 10, 10, 10)

dd <- genDataDensity(10000, dataDist = base_data, varname = "x1")

Plotting the generated data with the original density:

de <- density(base_data, n = 100000)
de <- data.table(x = de$x, y = de$y)

ggplot(data = dd, aes(x=x1)) +
  geom_histogram(aes(y = after_stat(count / sum(count))), binwidth = 1, fill = "lightblue", color = "black") +
  geom_line(data = de, aes(x = x, y = y), color = "red") +
  theme(panel.grid = element_blank()) +
  scale_y_continuous(name = "density\n")

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 823090c is merged into main:

• ✔️define_data: 17ms -> 17.2ms [-1.6%, +2.96%]
• ✔️dist_beta: 212ms -> 212ms [-0.65%, +0.92%]
• ✔️dist_binary: 8.76ms -> 8.82ms [-1.25%, +2.53%]
• ✔️dist_binomial: 13.2ms -> 13.2ms [-1.39%, +1.4%]
• ✔️dist_categorical: 58.3ms -> 58.3ms [-0.69%, +0.93%]
• 🚀dist_exponential: 9.11ms -> 8.93ms [-3.56%, -0.29%]
• ✔️dist_gamma: 16.5ms -> 16.6ms [-1.85%, +2.95%]
• ✔️dist_mixture: 228ms -> 228ms [-2.54%, +2.92%]
• ✔️dist_normal: 10.8ms -> 10.7ms [-2.55%, +0.95%]
• ✔️gen_all_dists: 48.8ms -> 49.2ms [-1.6%, +3.3%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[kgoldfeld]

@assignUser Let me know if you want to take a look before I go ahead and merge this.

[assignUser]

Cool, great work on the tests too!