Improved family-wise error rate control in multiple equivalence testing.

Area of Science:

• Agricultural Science
• Statistical Science
• Food Safety Science

Background:

• Equivalence testing is crucial for food and feed safety assessments, enabling market approval by demonstrating crop characteristic equivalence.
• Current univariate equivalence tests applied per analyte lack multiplicity correction, increasing Type I error rates (false claims of equivalence) when assessing multiple characteristics.

Purpose of the Study:

• To evaluate and compare the power of different familywise error rate (FWER) controlling methods for equivalence testing in food safety.
• To identify more powerful alternatives to Hochberg's method for managing multiple comparisons in equivalence assessments.

Main Methods:

• Comparison of Hochberg's method with other FWER-controlling procedures, including Hommel's method and an adaptive Bonferroni method.
• Application of these methods to two real-world compositional datasets.
• Assessment and comparison using simulated data to evaluate performance under various scenarios.

Main Results:

• Hommel's method is demonstrated to be at least as powerful as Hochberg's method.
• An adaptive Bonferroni method, utilizing an estimator of non-equivalent characteristics, frequently shows substantially greater power than Hommel's method.
• The adaptive Bonferroni method is particularly advantageous in food safety contexts where a high proportion of true equivalences is anticipated.

Conclusions:

• Standard equivalence testing practices in food safety may be overly conservative due to uncorrected multiplicity.
• Adaptive Bonferroni and Hommel's methods offer improved statistical power for FWER control in multi-analyte equivalence testing.
• These advanced methods enhance the accuracy and efficiency of safety assessments for new food and feed products.