There's no reason that the signed rank can't be used for samples above a certain sample size, or below a certain size.

For small sample sizes, a software implementation may use an exact test for the determination of the p-value.

However, a sample size of 6 pairs is the minimum to get a p-value below 0.05. You can see this if you compare the results for:

A = (1,2,3,4,5)

B = (2,4,6,8,10)

and

C = (1,2,3,4,5,6)

D = (2,4,6,8,10,12)

But try

wilcox.test(rep(1,5))

Sal Mangiafico . Albeit you get a warning, and there isn't much point doing a hypothesis test with n=5.

Daniel Wright , interestingly, for that example, the p-value shifts across the magic "0.05" level depending on whether the exact approach is used or the asymptotic approach (even with continuity correction). e.g.,

A = c(1, 1, 1, 1, 1)

library(exactRankTests)

wilcox.exact(A, exact = TRUE)

wilcox.test(A)

I saw your note Sal Mangiafico and was curious. And you can use a t test with 2 cases and get p

Daniel Wright , it's because the t statistic is based on the ratio of the difference of the means relative to the variance in the data (roughly speaking). On the other hand, because the signed rank test uses the ranks, it doesn't matter for your example if the data were (2,2,2,2,2) or (100,100,100,100,100).

Thank kindly at these insights Daniel Wright and Sal Mangiafico