ESTIMATION OF TOTAL TIME ON TEST TRANSFORMS FOR STATIONARY OBSERVATIONS

By proving Chibisov-O'Reilly-type theorems for uniform empirical and q uantile processes based on stationary observations, we establish a non parametric large sample estimation theory for total time on test trans forms. In particular, we obtain weak approximations for total time on test transforms also under the assumption of positively associated dep endence, a kind of dependence that is encountered in many practical li fe testing situations. We derive similar asymptotic results for mixing sequences as well, another and often used structure of dependence for sequences.