Two-Tailed Bootstrap P-Value

The share of bootstrap estimates on the far side of a null value, doubled: with `lower` the proportion of estimates at or below `null_value - epsilon` and `upper` the proportion at or above `null_value + epsilon`, the p-value is `2 * min(lower, upper)`. Estimates within `epsilon` of the null value count on neither side. Missing estimates are ignored.

Usage

boot_p_value(estimates, null_value = 0, epsilon = 1e-05)

Arguments

estimates

Numeric vector of bootstrap estimates.

null_value

The value under the null hypothesis (default 0).

epsilon

Tolerance around the null value (default `1e-5`).

Value

A number between 0 and 1; `NA` when there are no estimates. Estimates that equal the null value exactly count on both sides when `epsilon` is 0, and the result is capped at 1. A p-value of 0 means no estimate lies on the other side of the null: it says the p-value is below about 2 / (n + 1), not that it is zero.

Examples

boot_p_value(c(0.5, 0.7, 0.9, 1.1, -0.2))
#> [1] 0.4