Hi, I am working on the discriminating power of a test result of a class. I have only searched the internet and youtube to learn more about this.

There are 54 students in my class. Each student answers 10 questions. Each questions carries at most 10. Total mark is 100.

First, I want to know if discriminating power is related to the difficulty index and discrimination index. Here is what I could understand:

1/ the total score of the whole class needs to be sorted in descending.

2/ Divide 54 students into 3 groups. Focus on top 27% (15) and bottom 27% (15).

my problems are:

1/ the scores of students are not 1 or 0. Each score is between 0 and 10. So we are not calculating # correct as in the example online. Must I add a value with only 1 or 0?? and is there suggestion on how to define this value??

2/ Do I need to use Point Biserial correlation each Q and total score?? and use the p-value of the outcome?? I am still not sure if I can get the discriminating power from this.

3/ Is the correlation coefficient between each Q and the total mark useful??

There are 54 students in my class. Each student answers 10 questions. Each questions carries at most 10. Total mark is 100.

First, I want to know if discriminating power is related to the difficulty index and discrimination index. Here is what I could understand:

1/ the total score of the whole class needs to be sorted in descending.

2/ Divide 54 students into 3 groups. Focus on top 27% (15) and bottom 27% (15).

my problems are:

1/ the scores of students are not 1 or 0. Each score is between 0 and 10. So we are not calculating # correct as in the example online. Must I add a value with only 1 or 0?? and is there suggestion on how to define this value??

2/ Do I need to use Point Biserial correlation each Q and total score?? and use the p-value of the outcome?? I am still not sure if I can get the discriminating power from this.

3/ Is the correlation coefficient between each Q and the total mark useful??

Last edited: