Linear Statistical Models
Linear models form the core of classical statistics, and S provides extensive facilities to fit and investigate them. These work with a version of the Wilkinson-Rogers notation (Wilkinson & Rogers, 1973) for specifying models which we discuss in Section 6.2.
KeywordsModel Matrix Data Frame Heart Weight High Order Interaction Diagnostic Plot
Unable to display preview. Download preview PDF.
- 1.The Sex coefficient appears to be significant but this is misleading as this term is marginal to log(Bwt) %in% Sex and may be modified by a change of origin.Google Scholar
- 2.Marginal terms are sometimes removed in this way in order to calculate what are known as “Type III sums of squares” but we have yet to see a situation where this makes compelling statistical sense.Google Scholar