Comparing traditional and continuous norming models for neurobehavioral assessments: A simulation study

Developing accurate test norms is crucial to developmental disability practice but requires substantial resources. Continuous norms show promise relative to traditional norms, yet there is limited guidance on optimal methods and minimum sample sizes. This simulation study compared traditional norming with continuous norming models across 96 data conditions varying in sample size (500−1500), age trajectories (ages 2–22), variance patterns, and skewness. Generalized Additive Models for Location, Scale, and Shape (GAMLSS) that approximated the data generating conditions tended to be the best fitting models. Best fitting GAMLSS models showed closer fit to the data at N = 500 than traditional windowing norms at N = 1500 with the largest GAMLSS performance gains occurring by N = 750. Real-world neurobehavioral data supported the need for an iterative model selection process when developing continuous norms. Test developers can achieve high norming accuracy with smaller samples using a GAMLSS model selection process, potentially reducing costs while improving precision.