explore_stephenson_s_finite: Explore power and precision across a range of s values

In li-xinran/RIQITE: Randomization Inference for Quantiles of Individual Treatment Effects

Explore power and precision across a range of s values

Description

Conduct power simulation against specific finite-sample dataset to see how power changes as a function of s.

Usage

explore_stephenson_s_finite(
  s = c(2, 5, 10, 30),
  Y0,
  tau,
  p_tx,
  R = 100,
  percentile = 1,
  k.vec = NULL,
  alpha = 0.05,
  c = 0,
  quantile_n_CI = NA,
  alternative = "greater",
  stat.nulls = NULL,
  nperm = 1000,
  targeted_power = TRUE
)

Arguments

s	Numeric list of s values to test using the Stephenson rank.
Y0	Either a name of function to generate Y0s, or an empirical distribution represented as a list of values.
p_tx	Proportion of units to be treated.
R	Number of simulation iterations to do.
percentile	For targeted power, what percentile to target.
k.vec	Which quantiles to examine (if targeted_power = FALSE).
alpha	Significance level of the test.
c	Threshold bound for treatment impacts for hypothesis testing.
quantile_n_CI	What quantile of the distribution of lower confidence limits for the number of significant units to calculate. NA (the default) means calculate mean instead of median. 0.5 would report the median size of the confidence interval for number of significant units.
alternative	A character takes value "greater" and "less", indicating the direction of alternative hypotheses.
stat.nulls	Optional list of the precomputed null distributions of the statistics.
targeted_power	TRUE means focus on passed percentile and calculate power for that specific quantile. FALSE means calculate simultenaous confidence intervals for all quantiles, leaving power caculation to post-simulation exploration.

li-xinran/RIQITE documentation built on July 1, 2023, 6:58 p.m.