Sankhya: The Indian Journal of Statistics

2000, Volume 62, Series A, Pt. 2, 223--235

ACCELERATED LIFE TESTING MODEL BUILDING WITH BOX-COX TRANSFORMATION

By

PEIHUA QIU, University of Minnesota, Minneapolis

and

CHRIS P. TSOKOS, University of South Florida, Tampa

SUMMARY. In accelerated life testing, the nominal life time is often related to stress levels by an acceleration equation. Three particular models that have been used frequently in the past are the power law model, the Arrhenius model and the Eyring model. In this paper we suggest choosing a model from a model family which includes the three particular models as special cases. This family is defined by a Box-Cox transformation on the stress variable. There are two benefits to use this proposal: (1) model fitting could be treated in an unified way; and (2) the fitted model is more robust to model assumptions. We demonstrate this method with some numerical examples.

AMS (1991) subject classification. Primary 62N05; secondary 62F25.

Key words and phrases. Accelerated life testing, acceleration equation, Box-Cox transformation, residual sum of squares.

Full paper (PDF)

This article in Mathematical Reviews