Coefficient estimates are presented for regression analysis with missing values of the independent variables. It is assumed that the missing data are not missing completely at random. The six proposed methods are, namely, complete-case analysis, available-case methods, least squares, maximum likelihood, Bayesian methods and multiple imputations. The case of missing data in one independent variable is first tackled, and extensions to more general cases are given. Lastly, the Bayesian methods and the multiple imputations are considered the most promising.