Ch. 7: Test Construction
I. Introduction
A. Test construction is a process of item selection from an item bank for a specified
purpose.
B. In CTT item and test indices are variant over groups.
1. The index group and the target group should be matched in their ability
(school start, end, and year).
2. Not all experimental (filler) items can be included in one test. No
guarantee for equivalent indices.
3. Standard error of measurement (SEM) and test reliability are general
indices for a whole test. We cannot have SEM and reliability for a specific
ability range.
4. Contribution of each item to a test cannot be isolated. Each item is related
to other items in a test.
C. In IRT, item parameters are invariant.
1. Problems 1 – 4 in CTT are resolved in IRT.
2. and b are on the same scale (0, 1). Select appropriate items for a
specific ability range.
3. Item information of each item is independent of other items in a test.
SE( ) can be set and items with large Ii( ) can be selected for a specific
purpose of a test.
II. Procedure (Lord, 1977)
A. Develop a target information function for a test (p. 104, f.7.1).
1. Decide ability range (e.g. -2
2
).
2. Decide SE( ) =
)(
1
I
(e.g., .50, .33, .25).
B. Select items from the item bank to match the target information function. Pay
special attention to the tails of the ability distribution.
C. After each item is added to the test, calculate the test information function.
D. Continue the item selection procedure until the test information function is similar
to the target information function.
E. We may set a criterion for the number of items for the test as an alternate for the
Docsity.com