L’EMbeDS DS^3: Recent advances in equivalence testing - with Luca Insolia
The DS3 organizing committee, alongside the L'EMbeDS Department of Excellence of the Sant'Anna School for Advanced Studies launches a new cycle of seminars devoted to exploring frontier research in data science and its applications across disciplines.
The L'EMbeDS Data Science Seminar series hosts international scholars and experts to discuss cutting-edge methodologies and their implications for economics, social sciences, and beyond.
For announcements and further information please visit our web page or join to our Google Group.
The first seminar will feature Luca Insolia (University of Geneva), who will present a talk entitled: “Recent Advances in Equivalence Testing”
ABSTRACT:
Equivalence testing aims to assess whether an effect of interest, such as the difference in means between two treatment outcomes, lies within a predetermined region of practical equivalence. Assessments of equivalence arise in several domains, such as economics, psychology and engineering, and they play a key role in the pharmaceutical sciences (e.g., for the regulatory approval of generic drug products). The standard method to assess equivalence, the Two One-Sided Tests (TOST) procedure, is known to be overly conservative, resulting in a loss of statistical power. To mitigate this problem in the context of averages and quantiles equivalence testing, we propose a general family of finite-sample adjustments for the TOST.
These adjustments are designed to ensure that the theoretical test size matches the nominal significance level, leading to uniformly more powerful test procedures compared to the traditional TOST. We establish the theoretical properties of these adjusted procedures, and empirically demonstrate their superior performance through extensive simulations covering a range of challenging scenarios.
These include multivariate assessments with relatively small sample sizes, unknown and heterogeneous variances, and various correlation structures. We illustrate the practical relevance of our approach through examples related to economics and the pharmaceutical sciences.