PRM106 - Calculating Rasch-Based Person Parameters for Individual Items

Abstract

The Rasch model is the most constrained of the item response theory (IRT) models and is often used for analyzing categorical data based on (a) the respondent's trait levels (e.g., their disease severity) and (b) the item difficulty (i.e., how unlikely it is for a given disease characteristic to be reported). Rasch data are calculated at the domain level, and may sometimes be appropriate for a total score. We can examine more specific information by looking at individual scores on the entire domain or item-level data for all individuals at once. There are times, however, when one may wish to examine scores for each individual on a single item. For example, one may wish to examine the influence of changing a raw-score scale range by comparing it to Rasch data for the same item. Unfortunately, none of the current IRT software permits these calculations. The goal of the current study was to refine the process for such analyses, thereby creating an Excel worksheet, and make them available for others. In 1998, Linacre [1] published the details of how to calculate individual-level Rasch parameters for each item of a domain. As the formulae are cumbersome and lengthy, an easy to use Excel worksheet was created. The Excel worksheet was first reviewed by members of our team and then tested with data from a depression patient-reported outcome measure. Qualitative and quantitative review of the worksheet found that, beyond being easy to use, it was an accurate and appropriate representation of the Linacre formulae. Programming details and Excel layout are provided in the poster presentation and will be made available from the first author. Linacre, J.M., Estimating Rasch measures with known polytomous item difficulties. Rasch Measurement Transactions, 1998. 12: p. 638.