HLMdiag: An Implementation of Diagnostics for Hierarchical Linear Models in R

• Activity Number: 612
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Graphics
• Author(s): Adam Loy*+ and Heike Hofmann
• Companies: Iowa State University and Iowa State University
• Keywords: R package ; residual analysis ; graphics for hierarchical linear models ; deletion diagnostics ; level-based residuals

Abstract

There are numerous familiar diagnostic measures for the linear regression model that enable us to check model assumptions and identify influential observations that might distort parameter estimates, predictions, and the precision of both. We describe how these techniques can be adapted for the more complex family of Hierarchical Linear Models (HLM). Hierarchical models do not assume independence between data points. This additional dependence structure has to be considered in the diagnosis of the model, too. In particular, this means that observations can be influential at multiple levels of a model. In this paper we discuss the implementation of residual analysis and deletion diagnostics for checking the model assumptions and detecting influential points in the hierarchical linear model in the R package HLMdiag, including application to a case study.