Methodological considerations when establishing reliable and valid normative data: Canadian Longitudinal Study on...

Area of Science:

• Neuroscience
• Psychometrics
• Gerontology

Background:

• Normative data are crucial for interpreting cognitive test scores.
• Traditional methods risk small cell sizes, impacting reliability.
• Regression corrections for demographic factors are increasingly used.

Purpose of the Study:

• To compare regression correction methods for demographic covariates in cognitive normative data.
• To assess the impact of covariate choice (age, sex, education) on normative data reliability and validity.
• To explore measurement invariance across different modeling approaches.

Main Methods:

• Utilized data from the Canadian Longitudinal Study on Aging (CLSA).
• Administered a brief, telephone-based cognitive battery.
• Explored measurement invariance for sex and education in full and hybrid regression models.
• Excluded participants with neurological conditions (N=12,350 English, N=1,760 French).

Main Results:

• Hybrid models supported measurement invariance for sex and education.
• Full regression models did not support measurement invariance.
• Reliability, measured by precision (95% inter-percentile range), was higher in full regression models.
• Precision differences were negligible in the larger English sample.

Conclusions:

• Presents normative data for a remotely administered neuropsychological battery.
• Hybrid models best mitigate measurement bias while maintaining precision.
• Large sample sizes are essential for reliable normative data, particularly for smaller subgroups.
• The study provides a syntax file for the resulting normative data.