Do I have to test normality?

ger95

New Member
#1
I have an experiment contains 10 students and 2 experts. I want to test if the average of students evaluation for the 1st question is similar to experts average. the question answer is scale from 1~5.
- do I have to test if students answer is normally distributed? you should know the students are the whole students in the class and they are not random sample.
- I plan to do one sample t-test for comparing students median and experts median.

thanks in advance
 

Karabiner

TS Contributor
#2
- do I have to test if students answer is normally distributed?
You probably have an ordinal scale, so normality isn't an issue.
And with n=10, any normality test has too small statistical
power to be useful. If you sticked to nonparametric tests, you'd
be on the safe side.

- I plan to do one sample t-test for comparing students median and experts median.
t-tests are for means, not for medians. You could do a
Mann-Whitney U-test, although I don't know
whether it is at all possible to reach the usual
p < 0.05 criterion if one of the groups has only
n=2.

With kind regards

K.
 

ger95

New Member
#3
thanks Karabiner, and one thing more, if I have more than one question such as 20 and experts answer all and each student answer 5. do I have to do Kappa for the experts or I can just take the average.