We consider a model for a response variable as the sum of a linear function of a predictor
variable and independent random variation. This is known as the linear regression model.
We discuss the linear regression model in the particular context of a bivariate normal
distribution and relate it to the correlation coefficient. The linear model assumes the
predictor variable is measured without error. We consider a measurement error model for
cases when this assumption is unrealistic. Some functional relationships can be transformed
to linear relationships, and one such example is presented.
|