these commands in MTB show the effect of one outlier
on the t test and the wilcoxon test.
choose a sample size of n=20 from a normal(1,1) distr.
test the hypothesis H0: mean=0 .
with this data having mean=1, it should be rejected.

#########################
random 20 c1 ;
normal 1 1 .
sort c1 c1     #order the data for later use
#
onet c1 ;
test 0 .
Wtest  0  c1 ;
alternative 0 .
# 
let k1 = c1(20)
print k1
let c1(20) = c1(20) + 3     #make the largest obs larger (outlier)
#
# now repeat the last 5 command lines to do the tests again
# copy-paste the commands to do it
# repeat the tests, maybe 5 to 10 times
# make a table of your results giving: value of c1(20) and the 2 pvalues