Testing for Hidden Additivity in Factorial Experiments Using the Hiddenf Package in R

• Activity Number: 613
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Christopher Franck* and Jason Osborne
• Companies: Virginia Tech and North Carolina State University
• Keywords: interaction effects ; latent variables ; hidden additivity ; non-additivity

Abstract

Unreplicated two-way factorial experiments enjoy prevalent usage across many scientific disciplines due to their simplicity and low cost. One drawback of this design is that standard F tests to assess interaction effects in these experiments are unavailable. The study of interaction is fundamental to statistical inference, so the development of software to support existing methods to detect non-additivity in this class is an important goal. This talk describes hiddenf, an R package that provides functionality to diagnose non-additivity in two way unreplcited experiments. The package includes two classes of functionality. First, the package conducts 5 available tests for non-additivity proposed between 1948 and 2014. Second, the package provides expanded diagnostics for the detection of hidden additivity. Hidden additivity is a form of interaction that occurs when latent groupings of levels in one factor interact with the levels of the other factor. Functionality is demonstrated on 15 sets of data, many of which have been investigated in previous articles relevant to this topic. The hiddenf package is available on the R Comprehensive Archive Network