Disentangling individual differences in cognitive response mechanisms for rating scale items: A flexible-mixture...

Area of Science:

• Psychometrics
• Statistical Modeling
• Behavioral Science

Background:

• Item Response Theory (IRT) models analyze rating-scale data but often assume uniform response strategies.
• Heuristic response strategies, such as response styles (RS), can influence accuracy but are not always adequately modeled.
• Traditional IRTree models address response styles but assume all respondents use the same strategy, potentially causing inaccurate inferences.

Purpose of the Study:

• To introduce the mixture IRTree (MixTree) model, which accommodates individual differences in response strategies.
• To improve the accuracy of inferences from rating-scale items by accounting for response style heterogeneity.
• To identify underlying sources of variation in response processes among individuals.

Main Methods:

• Developed the mixture IRTree (MixTree) model, assigning participants to latent classes with distinct response processes.
• Incorporated varying weights for trait and extreme response style (ERS) scores based on class membership.
• Utilized simulation studies for model validation and empirical data analysis to identify latent classes.

Main Results:

• Simulation studies confirmed MixTree's ability to accurately recover latent classes and model parameters.
• Empirical data analysis revealed two distinct latent classes of respondents.
• One identified class was associated with trait-driven response mechanisms, while the other was linked to response style-driven mechanisms.

Conclusions:

• The MixTree model offers a more accurate approach to analyzing rating-scale data by acknowledging response strategy heterogeneity.
• Individual differences in response processes, such as trait-driven versus response-style-driven mechanisms, can be identified.
• This approach enhances the validity of inferences drawn from psychometric assessments.