Improvement of Norm Score Quality via Regression-Based Continuous Norming

Area of Science:

• Psychometrics
• Statistical modeling
• Psychological assessment

Background:

• Psychometric test interpretation relies on norm scores.
• Conventional norming methods involve ranking procedures and compiling norm tables.
• The accuracy of conventional norming can be influenced by sample size and data characteristics.

Purpose of the Study:

• To compare semiparametric continuous norming (SPCN) with conventional norming methods.
• To evaluate the performance of different norming approaches using item response theory simulations.
• To identify potential errors and limitations in conventional norming procedures.

Main Methods:

• Simulated psychometric test data using item response theory with varying item properties.
• Modeled norm scores using random samples with conventional ranking and SPCN.
• Cross-validated norms with a large representative sample (N=840,000) and computed norming errors.
• Repeated the simulation and validation process 90,000 times.

Main Results:

• Both SPCN and conventional norming improved with larger sample sizes.
• SPCN achieved optimal results with significantly smaller samples than conventional norming.
• Conventional norming exhibited poorer data fit, more age-related errors, and a higher number of missing values in norm tables.
• Data fit in conventional norming varied with age bracket granularity, questioning sample size recommendations.

Conclusions:

• SPCN is a more efficient and accurate method for psychometric test norming.
• Conventional norming methods have limitations regarding data fit, error rates, and completeness.
• Test norms should be derived from statistical models of raw score distributions rather than conventional ranking procedures.