Abstract : Diagnostics based on robust R-estimates of regression coefficients are developed. These methods are not as sensitive to influential points as least squares diagnostics. In data sets with several influential points, diagnostics based on a robust fit have a greater chance of detecting interesting cases for further inspection. Robust analogues of the internal and external t statistics, DFFITS, DCOOK, and DFBETAS are developed and illustrated on two date sets. Keywords: Linear models, Statistical data, Tables data.