Dear Yalda Zolghadri,

Since you are comparing the grades (AB,C) among the two classes of a school, its a attribute data, and you need to apply chi-square test to check there are statically significant or not.

In Graph -Pad, go to the contingency option.

Make 3 x 2 contigency table (3 rows, 2 colums)

Grade Class-1 Class-2

A

B

C

Apply Chi-square (and Fisher's Exact test).

Regards

Vishwanatha

Thank you so much Vishwanath shastry G.L.

It was very helpful.

Best,

Y

4u

with Prism 7

Thank You Abolfazl Ghoodjani

This was awesome!

Best,

Y

A chi-square test or Fisher's test will not answer the question asked, which was Which class performed better?.

As an illustration of this, imagine Class 1 had 10 A, 30 B, and 10 C, while Class 2 had 20 A, 10 B, and 20 C. A chi-square test will return a significant result, but it's clear that both classes have a B average. This is because these tests compare the distributions or proportions of unordered categories.

In order to answer the question about which class performed better, the grades need to be treated as at least ordinal. A common approach would be to use a Wilcoxon-Mann-Whitney test. A Cochran-Armitage test is also applicable.

Following any one of these simple approaches (chi-square, Fisher, WMW, CA), may lead to incorrect conclusions with having different tests and missing data. Imagine that one test was much more difficult that the others, and several students in one class only didn't take that test. In that case, that class would have an unfair advantage in their scores. The same kind of argument could be made for a student who scores very poorly in general, and then takes only one test rather than all three. The solution here is to use a more complex model, presumably ordinal regression), that takes into account the differences among the tests and among the students. This is probably excessive for the problem being studied, but it is a good idea to check the average grade for each exam and for each student to be sure there isn't any obvious bias introduced by the missing data.

The grade ABC is not nominal but rather nominal. You may consider Mann Whitney test or Wilcoxon rank