Using SAS PROC NLMIXED to fit item response theory models.

Area of Science:

• Psychometrics
• Educational Measurement
• Statistics

Background:

• Item response theory (IRT) is widely used for developing and assessing measurement instruments like tests and questionnaires.
• Existing statistical packages lack built-in routines for IRT modeling, hindering its integration into academic curricula.
• Generalized linear mixed effects models offer a flexible framework that encompasses IRT models.

Purpose of the Study:

• To demonstrate the flexibility and generality of using Statistical Analysis System (SAS) for estimating IRT model parameters.
• To illustrate the implementation of various IRT models for different response types using real data.
• To encourage the adoption of IRT modeling in quantitative courses by leveraging the widespread availability of SAS.

Main Methods:

• Utilized Statistical Analysis System (SAS) software for parameter estimation in IRT models.
• Applied generalized linear mixed effects models as a unifying framework for IRT.
• Demonstrated implementations for dichotomous, polytomous, and nominal response data using real-world examples.

Main Results:

• Successfully estimated IRT model parameters using SAS, showcasing its capability for complex measurement analyses.
• Provided practical examples of applying diverse IRT models within the SAS environment.
• Confirmed that SAS can effectively handle various response formats common in surveys and tests.

Conclusions:

• SAS provides a powerful and flexible platform for implementing a wide range of Item Response Theory models.
• The integration of IRT modeling into SAS facilitates its application in research and education.
• This approach can enhance the teaching and application of advanced psychometric methods in quantitative courses.