This dataset is similar to that from "Uni Class A" in that it is from a North American university and involves an instructor-created multiple-choice test.  There were 34 test items responded to by 132 students.

 

We'll replicate the approach applied in the analysis of "Uni Class A", using Lertap's RSA routine to look for an indication of possible cheating.  You will see that there were more suspects in this class, with results in some disagreement with those from the SCheck program.

 

A copy of the dataset, as an Excel workbook, may be downloaded here.

 

The Freqs report was as clean as that in Uni Class A.  In fact, it was somewhat cleaner, with no asterisks, and even fewer unanswered questions.

 

The distribution of test scores is shown below:

 

 

The distribution is fairly similar to that from Uni Class A.  The average score was 52.5% (17.85 out of a maximum possible score of 34), standard deviation 6.1 (compared to 4.5 in Uni Class A results).

 

   

 

This test's coefficient alpha value was higher; there was only one item, I29, with discrimination below 0.10, and only one really easy item, I21.

 

Page forward for RSA results.