Simultaneous comparisons of treatments versus control (Dunnett-type tests) for location-scale alternatives

TL;DR

This paper proposes and compares several Dunnett-type tests for simultaneous treatment versus control comparisons, considering both location and scale effects, supported by simulations and real data examples.

Contribution

It introduces new approaches for Dunnett-type tests that simultaneously assess location and scale effects, expanding traditional methods focused only on location.

Findings

New tests perform well in simulations

Real data examples demonstrate practical applicability

R-code implementation provided

Abstract

Commonly, the comparisons of treatment groups versus a control is performed for location effects only where possible scale effects are considered as disturbing. Sometimes scale effects are also relevant, as a kind of early indicator for changes. Here several approaches for Dunnett-type tests for location or scale effects are proposed and compared by a simulation study. Two real data examples are analysed accordingly and the related R-code is available in the Appendix.