Select by clicking on "Multiple Item ICCs" in the Graphs window or from the Graphs menu. This enables the display of multiple model and empirical item characteristic curves/item response functions ICCs/IRFs and item information functions IIFs on the same graph.
On the Graphs pull-down menu, please check "Display by Item"
Click on the "Model", "Empirical", and "Information" curves you wish to display. Click again on your selection to clear it. Click on "OK" to display the curves.
Buttons are described in Graph window. The x-axis can use absolute scaling relative to the latent trait (as in this Figure) or scaling relative to each item difficulty. Absolute scaling is in logits or your user-scaled units (USCALE=, UIMEAN=) or raw scores.
Displayed are the selected model and empirical ICCs and IIFs. A similar technique is used to display DIF item characteristic curves for non-uniform DIF.
Example: Example0.txt - Partial credit ICC's - ISGROUPS=0
"Graphs" window, "Multiple-item ICCs", "Select All Model ICCs", OK
When the curves are plotted "Measure relative to item difficulty", they all center on zero. The curves squash together.
Click on "Absolute x-axis". The curves are plotted "Measure on latent variable". Each curve centers on its own item difficulty. The curves spread out.
Example: We want to compare two ICCs |
|
title = "Compare ICCs" ni = 2 item1 = 1 name1 = 1 codes = 0123 ; 4 category rating scales isgroups=0 IAFILE=* 1 0 2 0 * SAFILE=* 1 0 0 ; anchor values for first rating scale 1 1 -0.8 1 2 -0.3 1 3 1.1 2 0 0 anchor values for second rating scale 2 1 -1.1 2 2 -0.25 2 3 1.3 * &END Scale 1 Scale 2 END LABELS 30 ; dummy data with 4 categories 03 11 22 |
The Andrich thresholds differ, but the ICCs are almost the same. |
