Table of Contents  |  Search Technical Documentation  |  References

NAEP Analysis and Scaling → Estimation of NAEP Score Scales → Checks for Violations of Assumptions → Comparisons of Empirical and Theoretical Item Response Functions → Items Showing Poor Item Fit

NAEP Technical DocumentationItems Showing Poor Item Fit

			Combining the Categories of Constructed-Response Items Treating Items as Different Across Assessment Years Deleting Items from Scales

Items with student responses that cannot be well described by Item Response Theory (IRT) models have empirical and theoretical item response functions (IRFs) that are not close to one another. Three actions that might need to be taken to assure that all items in a scale fit the models well are combining categories of constructed-response items, treating items as different for different assessment years, and deleting poor fitting items from the scale. Examples of items with response data that have been treated in these three ways provide guidelines for the recognition of poor fit. For these examples the symbols representing the empirical proportions do not overlay the theoretical probabilities that are represented by the continuous curves for at least part of the θ scale.

Printer-friendly Version