Sulla robustezza del test t di Student in presenza di un dato anomalo

A. Gambini
This paper is concerned with the influence analysis of outlying observation on the Student's t-test power function. This test is compared with the particular family of "robust procedures" consisting in the previous application to the data set of appropriate outlier identification and rejection rules and subsequent classical testing. We found out "break even sample sizes", starting from which the Student's t-test is more powerful than the "robust procedures" assumed as benchmarks.