Linear models, which are implicit in much of regression analyses, specify that the means of random variables are linearly related to values of covariates. Normality and equality of variancies of the random variables are usually assumed