51many different trait levels, we simply plot the probabilities that we have obtained.
The line connecting these probabilities reflects the item’s characteristic curve.
Figure 2.8 displays an item characteristic curve for a polytomous item. It shows
the highest probability of answering P0 (Strongly Disagree) is associated with the
lowest trait level. While P3 (Strongly Agree) corresponds with the highest trait
level.
Item information function can identify items that perform well or poorly. Low
information for one item may indicate that the item:
- Measures something different from other items in the scale
- Is poorly worded and need to be rewritten
- Too complex for the respondents
- Placed out of context in the questionnaire
Test Information
From the perspective of CTT, reliability was an important psychometric consider-
ation for a test. For example, we might compute coefficient alpha as an estimate of
the test’s reliability. An important point to note is that we would compute only 1
reliability estimate for a test, and that estimate would indicate the degree to which
observed test scores are correlated with true scores.
The idea that there is a single reliability for a particular test is an important way
in which CTT differs from IRT.
Fig. 2.8 Item characteristic curve for a polytomous item showing the probability of different
responses according to different trait levels
2 A Guide to PROMs Methodology and Selection Criteria