A tutorial on automatic post-stratification and weighting in conventional and regression-based norming of...

Area of Science:

• Psychometrics
• Statistical Modeling
• Developmental Psychology

Background:

• High-quality norm scores are vital for accurate individual diagnostics.
• Achieving representative normative samples is challenging due to resource limitations and complex stratification variables.
• Stratification variables are often collected post-hoc, necessitating post-stratification techniques.

Purpose of the Study:

• To introduce and demonstrate the application of raking (iterative proportional fitting) for enhancing normative sample representativeness.
• To outline the process of calculating weights, applying them to percentile estimation, and generating regression-based norm models.
• To provide a practical tutorial using the cNORM package in R.

Main Methods:

• Application of raking (iterative proportional fitting) for post-stratification of normative samples.
• Calculation of weights based on stratification variables (e.g., sex, ethnicity).
• Utilization of the cNORM R package for percentile estimation and continuous norm model retrieval.

Main Results:

• Demonstrated successful application of raking to a large, non-representative dataset (N=4542) of vocabulary development.
• Showcased the generation of regression-based norm models using weighted data.
• Validated raking as a method to simulate representativeness in normative samples.

Conclusions:

• Raking is an effective post-stratification technique to improve the quality and representativeness of normative samples.
• This method enhances the accuracy of norm scores, leading to better individual diagnostics.
• The cNORM package provides a robust framework for implementing raking in psychometric test construction.