Sliced Minimum Aberration Designs for Four-platform Experiments

Multivariate testing is a popular method to improve the layout of digital products such as a website and an application. Fractional factorial designs are usually used to perform online testing with large number of attributes. However, digital spaces present a new design challenge that doest not exist in the traditional experimental design literature: online testing is conducted across multiple platforms including desktops, tablets, smart-phones, and smart-watches. The existing experimental design literature does not offer precise guidance for such a multi-platform context. Sadeghi et al. (2016) introduced a statistical design framework to address the multi-platform feature of digital experiments. Sadeghi et al. (2016) also introduced a novel “sliced effect hierarchy” and formally defined sliced minimum aberration designs for two-platform experiments. In this paper, we extend the sliced minimum aberration designs for four-platform experiments. We define and provide guidance to construct sliced minimum aberration designs for a four-platform experiment using the concepts provided in Wu and Zhang (1993) and the underlying structure of multi-platform experiments provided by Sadeghi et al. (2016). We also tabulate sliced minimum aberration designs with 16, 32, and 64 runs for four-platform experiments.