Seemingly unrelated regression models generalize linear regression models byconsidering multiple regression equations that are linked by contemporaneouslycorrelated disturbances. Robust inference for seemingly unrelated regressionmodels is considered. MM-estimators are introduced to obtain estimators thathave both a high breakdown point and a high normal efficiency. A fast androbust bootstrap procedure is developed to obtain robust inference for theseestimators. Confidence intervals for the model parameters as well as hypothesistests for linear restrictions of the regression coefficients in seeminglyunrelated regression models are constructed. Moreover, in order to evaluate theneed for a seemingly unrelated regression model, a robust procedure is proposedto test for the presence of correlation among the disturbances. The performanceof the fast and robust bootstrap inference is evaluated empirically insimulation studies and illustrated on real data.
展开▼
