handle missing values when calculating confidence intervals

[simonpcouch]

Closes #520. :)

With this PR:

library(infer)

data <- data.frame(
   prop = seq(0, 1, length.out = 10),
   group = rep(c("a", "b"), each = 5L)
)

set.seed(1)
boot_dist <-
   data %>%
   specify(prop ~ group) %>%
   hypothesize(null = "independence") %>%
   generate(reps = 1000, type = "bootstrap") %>%
   calculate(stat = "diff in medians", order = c("b", "a"))

get_confidence_interval(boot_dist, .95)
#> Warning: 4 estimates were missing and were removed when 
#> calculating the confidence interval.
#> # A tibble: 1 × 2
#>   lower_ci upper_ci
#>      <dbl>    <dbl>
#> 1    0.278    0.889

^{Created on 2023-12-22 with reprex v2.0.2}

[simonpcouch]

Requesting your review so I don't forget to later, @mine-cetinkaya-rundel, but this review is not at all urgent. I'll be AFK the next couple weeks, so truly no rush.🙂☃️🎄

[mine-cetinkaya-rundel]

This looks good to me @simonpcouch! One question is whether we want this to be a warning or a message. If warning is consistent with other messaging in the package, then that's ok. Otherwise, I've been thinking about warnings as call to action and messages as call to attention. There is nothing the user could do to address this, it's just good for them to know, so it feels more like a message to me. But again, consistency with the rest of the package is probably more important.

[simonpcouch]

I think that's a good point. Without regard for consistency, I think I'd agree with you and go ahead and change that warning to a message. There is some precedent for warning when informing the user about missing values, though. :/

infer/R/specify.R

Line 78 in e5095f0

cli_warn("Removed {sum(!is_complete)} rows containing missing values.")

infer/R/calculate.R

Lines 35 to 39 in e5095f0

	#' In some cases, when bootstrapping with small samples, some generated
	#' bootstrap samples will have only one level of the explanatory variable
	#' present. For some test statistics, the calculated statistic in these
	#' cases will be NaN. The package will omit non-finite values from
	#' visualizations (with a warning) and raise an error in p-value calculations.

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.