Computer adaptive testing tutorial

This tutorial shows the statistics and logic behind computer adaptive testing (CAT). With CAT, each examinee receives a different sequence of items depending on which items are answered correctly. After each item is administered, a variety of statistics are computed in order to identify which new item will provide the most information about the examinee. Actual test items are used. A true CAT program is running behind the scenes.

The goal of CAT, like any testing program, is for the test to closely estimate the examinee's true ability - the score they would receive on a perfectly valid and perfectly reliable test. Three ways of expressing true ability are z-scores, the SAT scale (200-800 mean=500 sd=100), and percentiles.

To start this tutorial, you need to select a true ability. You can use either z-scores, the SAT scale, or percentiles. The software will start by assuming your ability is equal to that of the average student (in the absence of any information, this is the safest assumption and best starting point). We will then see if the software is able to accurately estimate the true ability you have provided.

True ability in z-score units
-1.5     -1.0     -.5     0     .5     1.0     1.5

True ability in SAT units
350     400     450     500     550     600     650

True ability in percentiles:
5th     10th     25th     50th     75th     90th     95th