inst/tinytest/test-statistic.R

In modelsummary: Summary Tables and Plots for Statistical Models and Data: Beautiful, Customizable, and Publication-Ready

mod <- list()
mod$OLS <- lm(am ~ drat, data = mtcars)
mod$Logit <- glm(am ~ qsec, data = mtcars, family = binomial())

# unavailable estimate or statistic
expect_error(modelsummary(mod, statistic = "bad"), pattern = "not available")
expect_error(modelsummary(mod, estimate = "bad"), pattern = "not available")

# std.error
raw <- modelsummary(mod, statistic = 'std.error', output="dataframe")
truth <- c('(0.434)', '(0.120)')
expect_equivalent(truth, unname(raw[[4]][c(2, 4)]))
truth <- c('(4.045)', '(0.228)')
expect_equivalent(truth, unname(raw[[5]][c(2, 6)]))

# p.value
raw <- modelsummary(mod, statistic = 'p.value', fmt = '%.6f', output="dataframe")
truth <- c('(0.000078)', '(0.000005)')
expect_equivalent(truth, unname(raw[[4]][c(2, 4)]))
truth <- c('(0.241402)', '(0.206028)')
expect_equivalent(truth, unname(raw[[5]][c(2, 6)]))

# conf.int
raw <- modelsummary(mod, statistic = 'conf.int', output="dataframe")
truth <- c("[-2.873, -1.099]", "[0.421, 0.909]")
expect_equivalent(truth, unname(raw[[4]][c(2, 4)]))
truth <- c("[-2.760, 13.501]", "[-0.784, 0.131]")
expect_equivalent(truth, unname(raw[[5]][c(2, 6)]))

# conf.int, conf_level = 0.99
raw <- modelsummary(mod, statistic = 'conf.int', conf_level = .99, output="dataframe")
truth <- c("[-3.181, -0.791]", "[0.336, 0.994]")
expect_equivalent(truth, unname(raw[[4]][c(2, 4)]))
truth <- c("[-5.070, 16.689]", "[-0.966, 0.259]")
expect_equivalent(truth, unname(raw[[5]][c(2, 6)]))


# issue 722: renaming statistics
mod <- lm(mpg ~ factor(cyl), mtcars)

tab <- modelsummary(
  output = "dataframe",
  mod, 
  estimate = c("$\\hat{\\beta}$" = "estimate"),
  statistic = c("Confidence Interval" = "[{conf.low}, {conf.high}]"),
  shape = term ~ model + statistic)
expect_equivalent(
  colnames(tab),
  c("part", "term", "(1) / $\\hat{\\beta}$", "(1) / Confidence Interval")
)

tab <- modelsummary(
  mod, 
  output = "dataframe",
  estimate = c("$\\hat{\\beta}$" = "estimate"),
  statistic = c("t-stat" = "statistic", "p-value" = "p.value"),
  shape = term ~ model + statistic)
expect_equivalent(
  colnames(tab),
  c("part", "term", "(1) / $\\hat{\\beta}$", "(1) / t-stat", "(1) / p-value")
)

tab <- modelsummary(
  mod, 
  estimate = c("$\\hat{\\beta}$" = "estimate"),
  output = "dataframe",
  statistic = c("Confidence Interval" = "conf.int"),
  shape = term ~ model + statistic)
expect_equivalent(
  colnames(tab),
  c("part", "term", "(1) / $\\hat{\\beta}$", "(1) / Confidence Interval", "(1) /  ")
)