Analysis of Statistical Interactions in Factorial Experiments. (19th May 2014)

• Record Type: Journal Article
• Authors: Kitsche, A.; Schaarschmidt, F.
• Is Part Of: Journal of agronomy and crop science. Volume 201:Number 1(2015:Feb.)
• Page Start: 69
• Page End: 79
• Publication Date: 2014-05-19
• Subjects: Agronomy -- Periodicals; Crop science -- Periodicals; 630.5
• DOI: 10.1111/jac.12076
• Languages: English
• ISSNs: 0931-2250

Abstract

<abstract abstract-type="main" id="jac12076-abs-0001"> <title>Abstract</title> <p>Two or higher‐order factorial designs are very common in agricultural and horticultural experiments. The evaluation of such trials by analysis of variance (<sc>anova</sc>) and the corresponding <italic>F</italic>‐tests for the interaction effects covers only a global decision concerning the presence of interactions. This study presents a straightforward method, which provides a more detailed analysis of interactions via multiple contrast tests. The presented approach takes both the structure of each factor and the research question into account by building user‐defined product‐type contrasts. Simultaneous inference for these user‐specified interaction contrasts that controls the overall error rate is available. In addition to adjusted P‐values, it is recommended to use simultaneous confidence intervals to present the magnitude, direction and the biological relevance of the interaction effects. The proposed method is demonstrated using two horticultural trials. Furthermore, the authors provide a collection of worked examples using the R (A Language and Environment for Statistical Computing, 2013, R Foundation for Statistical Computing, Vienna, Austria) add‐on package statint stored on github (<ext-link ext-link-type="uri" xlink:href="https://github.com/AKitsche/statint" xlink:type="simple" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/AKitsche/statint</ext-link>).</p> </abstract>