ABSTRACT
This chapter extends the previous one by turning to multiple linear regression with error in one or more of the predictors and possible error in the response. A number of the key points from the discussion of simple linear regression carry over to here but there are some new dimensions (pun intended) to the problem with more than one predictor. Most importantly, the nature of the bias in naive estimators is more complex. The other big change is notational, as we now express all of the models and methods in matrix-vector form. There are many books on multiple linear regression that can be consulted. Some relatively applied texts include Kutner et al. (2005), Montgomery and Peck (1992) and Griffiths et al. (1993), while more advanced treatments can be found in linear models books, such as Seber and Lee (2003), Ravishankar and Dey (2000), etc.
