Univariate comparisons given aggregated normative data

Area of Science:

• Neuropsychology
• Statistical Methods

Background:

• Normative comparison is key in neuropsychological assessment to identify cognitive deviations.
• Multiple testing in assessments can inflate the familywise error rate (FWER).
• Existing FWER correction methods often require multivariate normative data, which is frequently unavailable.

Purpose of the Study:

• To evaluate FWER correction methods using aggregated normative databases.
• To compare the performance of different correction methods in simulated aggregated data scenarios.
• To provide guidance on selecting appropriate correction methods for neuropsychological assessments.

Main Methods:

• A simulation study was conducted to mimic aggregated database conditions.
• Compared were: no correction, Bonferroni correction, maximum distribution, and stepwise approaches.
• Evaluated were familywise error rate (FWER) and statistical power to detect deviations.

Main Results:

• The stepwise approach demonstrated superior performance in FWER and power when the aggregated database was complete.
• The Bonferroni correction yielded the lowest FWER when significant amounts of data were missing.
• Performance varied based on data completeness within the aggregated database.

Conclusions:

• The stepwise approach is generally recommended for neuropsychological normative comparisons.
• The Bonferroni correction is the preferred method when dealing with substantial missing data in aggregated norm databases.
• Method selection should be guided by the characteristics of the available normative data.