On the analysis of accelerated life-testing experiments