D-optimal design for the Rasch counts model with multiple binary predictors

Area of Science:

• Statistics
• Psychometrics
• Generalized Linear Models

Background:

• The Rasch Poisson counts model and its generalized negative binomial extension are valuable for analyzing count data.
• Incorporating binary predictors into the difficulty parameter enhances model flexibility.

Purpose of the Study:

• To derive optimal designs for Rasch Poisson and generalized negative binomial count models with binary predictors.
• To develop locally D-optimal designs for efficient estimation of regression coefficients.

Main Methods:

• Specifying the Rasch Poisson and generalized negative binomial models as generalized linear models.
• Deriving conditions for locally D-optimal designs based on effect sizes.
• Applying design theory to count data models.

Main Results:

• Locally D-optimal designs were developed for the Rasch Poisson and generalized negative binomial models.
• Conditions for optimality were derived, linking design characteristics to effect sizes.
• The findings are applicable to broader Poisson regression models.

Conclusions:

• The derived optimal designs facilitate efficient parameter estimation in complex count data models.
• The study highlights the utility of generalized linear model frameworks for psychometric modeling.
• Future research should explore the application of these designs in general Poisson regression for psychological studies.