ABSTRACT
Chapter 11 focuses on increasing precision in settings where bias is not at issue, particularly within a randomized clinical trial. It defines a precision variable and shows how it reduces sampling variability in the estimated treatment effect by effectively subtracting variability in the outcome. It proves that even when a linear model including the precision variable as an extra additive term is an incorrect model for the expectation of the outcome given the treatment and the precision variable, the estimated coefficient of treatment is unbiased and generally more precise than it would be without including the precision variable. It also points out that the loss of efficiency is negligible when the precision variable is not in fact associated with the outcome. Examples and R code are also provided.
