Computes the exact sign test for two dependent (paired) samples, the
standard reference test for ordinal paired observations referenced by Munzel and
Brunner (2002). It is based on the qualitative within-pair differences only: under the
null hypothesis the number of positive differences among the non-tied pairs follows a
Binomial(m, 0.5) distribution (equivalently, the shift algorithm of
pairedRankTest applied with all absolute differences set to one). Tied
pairs are discarded. The more efficient rank-based alternative is
pairedRankTest.
Usage
pairedSignTest(x, y, alternative = c("two.sided", "greater", "less"))
Value
a list with components: nPositive (number of pairs with x > y),
nNegative (number of pairs with x < y), nNonTied (number of
non-tied pairs used by the test), p.value, method and alternative.
Arguments
x
a numeric vector of observations from the first condition.
y
a numeric vector of observations from the second condition, paired
element-wise with x.
alternative
a character string specifying the alternative hypothesis, one of
"two.sided" (default), "greater" (x tends to exceed y) or
"less".
Author
Lech Madeyski
References
Munzel, U. and Brunner, E. (2002). An Exact Paired Rank Test.
Biometrical Journal 44(5), 558-569.
# Munzel and Brunner (2002) PGI example: the exact sign test reported p = 0.023.pairedSignTest(MunzelBrunner02.PGI$week4, MunzelBrunner02.PGI$baseline)$p.value