Calculate coverage intervals and confidence intervals for the sample mean, median, sd, proportion, ...
Typically, these will be used within df_stats(). For the mean, median, and sd, the variable x must be
quantitative. For proportions, the x can be anything; use the success argument to specify what
value you want the proportion of. Default for success is TRUE for x logical, or the first level returned
by unique for categorical or numerical variables.
coverage(x, level = 0.95, na.rm = TRUE)
ci.mean(x, level = 0.95, na.rm = TRUE)
ci.median(x, level = 0.9, na.rm = TRUE)
ci.sd(x, level = 0.95, na.rm = TRUE)
ci.prop(
x,
success = NULL,
level = 0.95,
method = c("Clopper-Pearson", "binom.test", "Score", "Wilson", "prop.test", "Wald",
"Agresti-Coull", "Plus4")
)a variable.
number in 0 to 1 specifying the confidence level for the interval. (Default: 0.95)
if TRUE disregard missing data
for proportions, this specifies the categorical level for which the calculation of proportion will
be done. Defaults: TRUE for logicals for which the proportion is to
be calculated.
for ci.prop(), the method to use in calculating the confidence
interval. See mosaic::binom.test() for details.
a named numerical vector with components lower and upper, and,
in the case of ci.prop(), center. When used the df_stats(), these components
are formed into a data frame.
Methods: ci.mean() uses the standard t confidence interval.
ci.median() uses the normal approximation method.
ci.sd() uses the chi-squared method.
ci.prop() uses the binomial method. In the usual situation where the mosaic package is available,
ci.prop() uses mosaic::binom.test() internally, which provides several
methods for the calculation. See the documentation
for binom.test() for details about the available methods. Clopper-Pearson is
the default method. When used with df_stats(), the confidence interval
is calculated for each group separately. For "pooled" confidence intervals, see methods
such as lm() or glm().
When using these functions with df_stats(), omit the x argument, which
will be supplied automatically by df_stats(). See examples.
# The central 95% interval
df_stats(hp ~ cyl, data = mtcars, c95 = coverage(0.95))
#> response cyl c95_lower c95_upper
#> 1 hp 4 54.50 112.000
#> 2 hp 6 105.75 167.200
#> 3 hp 8 150.00 311.925
# The confidence interval on the mean
df_stats(hp ~ cyl, data = mtcars, mean, ci.mean)
#> response cyl mean lower upper
#> 1 hp 4 82.63636 68.57236 96.70037
#> 2 hp 6 122.28571 99.84850 144.72293
#> 3 hp 8 209.21429 179.78111 238.64746
# What fraction of cars have 6 cylinders?
df_stats(mtcars, ~ cyl, six_cyl_prop = ci.prop(success = 6, level = 0.90))
#> response six_cyl_prop_lower six_cyl_prop_center six_cyl_prop_upper
#> 1 cyl 0.1074469 0.21875 0.3718991
# Use without `df_stats()` (rare)
ci.mean(mtcars$hp)
#> lower upper
#> 121.9679 171.4071