Automation of likely outliers detection in linear mixed models
Date
2013
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
It is difficult to detect outliers in linear mixed models. The traditional way of
identifying outliers is to check whether there are any violations in model assumptions
by examining the normal QQ plot and the residual plot. A simulation approach
proposed by Schützenmeister and Piepho adds the objectivity in interpreting results of
the QQ and residual plot. Based on this simulation approach, a software tool is
developed to indentify potential outliers in linear mixed models automatically. In
addition, the performance of this approach is evaluated. This tool is user-friendly to
inexperienced analysts and open sourced.