Premium
Diagnostic procedures
Author(s) -
McKean Joseph W.,
Sheather Simon J.
Publication year - 2009
Publication title -
wiley interdisciplinary reviews: computational statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.693
H-Index - 38
eISSN - 1939-0068
pISSN - 1939-5108
DOI - 10.1002/wics.12
Subject(s) - outlier , computer science , statistical graphics , normality , statistical model , data mining , graphics , visualization , linear model , statistics , econometrics , artificial intelligence , machine learning , mathematics , computer graphics (images)
Diagnostic procedures are used to check the quality of a fit of a model, to verify the validity of the assumptions behind the model, and to find outlying and/or highly influential observations. Our discussion focuses on the linear model (the most widely used model). Much of our discussion, however, pertains to other models, as we show when we extend our discussion to mixed models at the end of the paper. The traditional fit, least squares (LS), can be severely impaired by just one outlier. So along with LS we present two robust fits and diagnostic procedures which explore the differences among the three fits. These comparisons generally find the outlying and influential cases. Armed with this methodology, we then proceed to discuss diagnostics that explore the quality of fit and verify the validity of the assumptions, including independent and identically distributed errors and normality. Copyright © 2009 John Wiley & Sons, Inc. This article is categorized under: Statistical Models > Linear Models Statistical and Graphical Methods of Data Analysis > Robust Methods Statistical Models > Fitting Models Statistical and Graphical Methods of Data Analysis > Statistical Graphics and Visualization