I have two group of 25 and 27 people. I want to compare them on a sex activity score that vary from 0 to 9.
My distribution is algedly normal, the skewness and kurtosis respect parameter, but most people have answers like 0 or 1. I have three people eith a score of 4, and one 6.
Using z score only the 6 is flag an outliers.
Doing my test-t with dose valu result in non-signifiant difference between the two group. But with the same data a Mann-Withney U is significant.
Some people suggest to me, that I should do my analysis with and without the outlier an later on my study report the presence of outlier and the possible limitation of such result.
I'm wondering is this a good way of doing this or should I change the value of the outliers and if so using wich score (+ 1 from the last normal valu, + sd).
Also is my distribution really normal even if most of the people have a score of 0 or 1 with some 4 4 and a 6.
My distribution is algedly normal, the skewness and kurtosis respect parameter, but most people have answers like 0 or 1. I have three people eith a score of 4, and one 6.
Using z score only the 6 is flag an outliers.
Doing my test-t with dose valu result in non-signifiant difference between the two group. But with the same data a Mann-Withney U is significant.
Some people suggest to me, that I should do my analysis with and without the outlier an later on my study report the presence of outlier and the possible limitation of such result.
I'm wondering is this a good way of doing this or should I change the value of the outliers and if so using wich score (+ 1 from the last normal valu, + sd).
Also is my distribution really normal even if most of the people have a score of 0 or 1 with some 4 4 and a 6.
