corrected.ztest: Looney and Jones corrected Z-test

In samvanderpoel/partial-pairs: Statistics for Partially Matched Samples

Description

corrected.ztest uses the Looney and Jones corrected z-test to obtain a p-value for a partially matched pairs test.

Usage

corrected.ztest(x, y, alternative = c("two.sided", "greater", "less"))

Arguments

x	a non-empty numeric vector containing some NA values
y	a non-empty numeric vector containing some NA values
alternative	specification of the alternative hypothesis. Takes values: "two.sided", "greater", or "less".

Details

Looney and Jones’s corrected Z-test is “corrected” in the sense that it adjusts the standard error of the difference of the two samples by accounting for the correlation between the $n_1$ paired observations. Under the null hypothesis, the resulting test statistic Z_corr has an asymptotic N(0,1) distribution.

If proper sample size conditions are not met, then corrected.ztest may exit or perform a paired or unpaired two-sample t.test, depending on the nature of the sample size issue.

If the variance of input data is close to zero, corrected.ztest will return an error message.

Value

p-value associated with the hypothesis test

References

Kuan, Pei Fen, and Bo Huang. "A simple and robust method for partially matched samples using the p‐values pooling approach." Statistics in medicine 32.19 (2013): 3247-3259.

Examples

In the following, the true means are not equal:

x = rnorm(400, 0, 1)
x[sample(1:400, size=75, replace=FALSE)] = NA
y = rnorm(400, 0.4, 3)
y[sample(1:400, size=75, replace=FALSE)] = NA
corrected.ztest(x, y, alternative = 'two.sided')

samvanderpoel/partial-pairs documentation built on Dec. 22, 2021, 10:14 p.m.