np_stat_test: Non-parametric test for any statistic

In eliobartos/misc: Miscellaneous functions

Description

Performs a non parametric test for any statistics by bootstraping. Combines all values and samples vectors of length a_data, b_data. Then calculates differences in statistics to get distribution of differences in statatistics. Then compares observed difference with that distribution.

Usage

np_stat_test(
  a_data,
  b_data,
  stat_fun = mean,
  n_sim = 1000,
  alternative = "two.sided",
  alpha = 0.05,
  plot = TRUE,
  ...
)

Arguments

a_data	Numeric vector of values from group a.
b_data	Numeric vector of values from group b.
stat_fun	Function that calculates desired statistic.
n_sim	Number of times to sample groups and calculate difference.
alternative	Setting up alternative hypothesis: "two.sided" meaning stat(A) != stat(B), "less" meaning stat(A) < stat(B), "greater" meaning stat(A) > stat(B)
alpha	statistical significance
plot	(logical) Plot the distribution of differences in mean if TRUE.
...	Any additional parameters that stat_fun needs.

Details

H0: No difference in statistics

Value

returns a named list containing:

diff	all the statistics differences
observed	observed statistics difference
n_sim	number of simulations ran
p_value	two sided p-value for the test

Author(s)

Elio Bartoš

Examples

x = rnorm(300, mean = 5, sd = 1)
y = rnorm(200, mean = 4, sd = 1.5)
y2 = c(y, 110)

np_stat_test(x, y, mean)
np_stat_test(x, y2, mean)
np_stat_test(x, y, mean, trim = 0.1)
np_stat_test(x, y, median)

eliobartos/misc documentation built on Oct. 8, 2021, 1:10 a.m.