Before starting the next topic, CAT, I want to repeat these matters:
Longer tests are more reliable.
As more items are added to a test, reliability increases and measurement error decreases.
In CTT the most common measure of reliability is coefficient alpha.
Measurement error, in CTT, is the SEM or the CSEM -- CSEM is better.
As measurement error decreases, the size of the 68% and 95% confidence intervals also reduces.
I will display two workbooks as reminders.
One is the screening test used by a medical school in Prague.
The other involves the Geology431 test, and theta scores obtained by using Xcalibre.