대응 2표본에 대한 윌콕슨의 부호순위 검정(Paired Comparison : Wilcoxon Signed Rank Test)

install.packages("BSDA")

install.packages("exactRankTests")

library(BSDA)

library(exactRankTests)

# H0 : mu_A equal mu_B
# H1 : mu_A not equal mu_B

# 1. Dependent-samples Sign-Test
pre.A <- c(5, 2, 4, 7, 4, 4, 8, 3, 7, 6)
post.A <- c(6, 5, 4, 9, 4, 6, 8, 5, 6, 7)
BSDA::SIGN.test(pre.A, post.A, md = 0, paired = TRUE, alternative = "two.sided")

# 결과물
Dependent-samples Sign-Test

data: pre.A and post.A
S = 1, p-value = 0.125
alternative hypothesis: true median difference is not equal to 0
95 percent confidence interval:
-2 0
sample estimates:
median of x-y 
-1

Achieved and Interpolated Confidence Intervals:

Conf.Level L.E.pt U.E.pt
Lower Achieved CI 0.8906 -2 0
Interpolated CI 0.9500 -2 0
Upper Achieved CI 0.9785 -2 0

# 결론 : 유의확률이 0.125이므로 귀무가설을 기각할 수 없다.

# 2. Wilcoxon Signed Rank Test
wilcox.test(pre.A, post.A, paired = TRUE, alternative = "two.sided")

# 결과물
Wilcoxon signed rank test with continuity correction

data: pre.A and post.A
V = 2, p-value = 0.04858
alternative hypothesis: true location shift is not equal to 0

# 결론 : 유의확률이 0.04858이므로 귀무가설을 기각할 수 있다.

# 3. Exact Wilcoxon signed rank test
exactRankTests::wilcox.exact(pre.A, post.A, paired = TRUE, alternative = "two.sided")

# 결과물
Exact Wilcoxon signed rank test

data: pre.A and post.A
V = 2, p-value = 0.0625
alternative hypothesis: true mu is not equal to 0

# 결론 : 유의확률이 0.0625이므로 귀무가설을 기각할 수 없다.

[출처] R을 이용한 비모수 통계학, 임동훈 지음, 자유아카데미, p41~44