Is it ok to here test statistical significance using permutation test?

Question

I'm measuring electric signals from humans, who I have put into two different categories A and B. I would like know whether there is some difference between the two groups. My total sample size is $N = 14$ (seven subjects in both groups). Now, I do something to the signals and in the end get one number for each subject. How to test whether the two groups are statistically different? I was advised to do a permutation test (how?) because I don't know anything about the distributions of the numbers I'm comparing. Makes sense?

Answer 1

Here is a two sample permutation test in R:

install.packages('exactRankTests')
library(exactRankTests)


x<-rnorm(7, mean=2,sd=1)
y<-rnorm(7, mean=0,sd=1)


perm.test(x,y)

x and y are your data. As an example I just generated values from a normal distribution. Permutation tests are a type of statistical analysis known as nonparametric methods. They do not rely on assuming a distribution for your data. This is good when you have small samples and you are not able to rely on large sample approximations (e.g. Central Limit Theorem). These methods are not as powerful as the parametric methods which assume distributions, but are more generally applicable.