800-213-1673
skip to the main content area of this page
ExamDeveloper

Statistical Analysis


Statistical analysis helps even the best exams

Once the exam has been administered to the first group of at least 30-50 candidates, certain basic statistics should be analyzed before providing the score results to the examinees. This step is vital to ensure that there are no unintended consequences with any of the questions. Below are several reasons to review the exam’s statistical performance.

One reason for reviewing exam statistics is to improve the performance of each question. Even the most thorough, knowledgeable, and experienced exam development panel will create questions requiring revision after reviewing the examinee performance. For example, some questions may be so difficult that no examinee could correctly answer it. In addition, a question might be so easy that everyone gets the question correct. Perhaps no examinee is choosing option "D" for a question, indicating that it is the easiest option to eliminate. Finally, a question might be answered correctly by all of the poor performing examinees, but not be the high performers. All of these cases may represent a problem with a question that would only be detected after viewing the statistical performance.

Another reason to review the statistical performance of an exam is to prevent bias against a particular group of examinees. There are many laws which prohibit discrimination against individuals based on race, gender, creed, age, or familiar status. Therefore, it is important to review the demographic characteristics of those that pass or fail the examination to ensure that it does not unfairly discriminate based on these factors.

A third reason to review exam statistics is to understand changes in the relevance of topics in the industry or in educational preparation. Consider the situation where a question, which has performed well for the past 3 years, has suddenly shown to be too difficult. One potential reason for this might be that the topic being tested is either not being taught in courses or is no longer used in the industry. If this case, this would suggest that the question be revised or removed from the exam.

Following this eight-step process is the best way to ensure the highest quality results. Although it is possible to create an exam by eliminating some of these steps, it is not advisable to do so. Considering the amount the money spent to administer and maintain the exam, it does not make sense to cut corners during the development process. Besides, the cost of one legal challenge of a poorly developed examination will far exceed the amount spent on proper methodology.


Back to Overview

How frequently should you analyze statistics?

To ensure an examination is performing properly, the statistical properties of each question should be evaluated frequently. At a minimum, the statistical performance of an exam needs to be reviewed every 200 examinees or quarterly whichever occurs first.   

In addition, an annual review of the content relevance and passing standard should be conducted to ensure the exam is meeting the needs of the organization.

Version 3 of ExamDeveloper will contain the tools needed to do the statistical analysis online and provide immediate feedback of the results. Currently, the Exam Design staff can provide a statistical report including questions to improve or replace, usually within three business days.