Modeling Response Style Using Vignettes and Person-Specific Item Response Theory.

Area of Science:

• Psychometrics
• Personality Assessment
• Psychological Measurement

Background:

• Survey responses are influenced by item characteristics, respondent traits, and response styles.
• Anchoring vignettes offer fixed rating targets for modeling response styles in personality and attitudinal data.
• Previous research shows promise, but questions remain regarding necessary vignette numbers, model accuracy, and criterion validity improvements.

Purpose of the Study:

• To investigate the number of vignettes and ratings needed for response style parameter identification.
• To compare the accuracy of various response style models.
• To determine if correcting for response style enhances criterion validity.

Main Methods:

• Utilized both simulated data and data from a population-representative sample.
• Modeled data using the Personality Inventory for DSM-5 (PID-5) measure of personality pathology.
• Employed an array of response style models, including the graded response model (GRM).

Main Results:

• Most response style models outperformed the GRM in simulations.
• All response style models were superior to the GRM when using observed data.
• Correcting for response style had a small, yet sometimes significant, impact on predicting self-reported social dysfunction.

Conclusions:

• Response style models incorporating anchoring vignettes are more accurate than traditional models like the GRM.
• Adjusting for response styles in personality assessments can improve the prediction of real-world outcomes, such as social dysfunction.
• Further research is needed to optimize the number of vignettes for reliable response style parameter estimation.