We propose a new robust estimator of the regression coefficients in a linear regression model. The proposed estimator is the only robust estimator based on integration rather than optimization. It allows for dependence between errors and regressors, is -consistent, and asymptotically normal....