Progressive stress accelerated life tests under finite mixture models

In this paper, progressive stress accelerated life tests are considered when the lifetime of a product under use condition follows a finite mixture of distributions. The experiment is performed when each of the components in the mixture follows a general class of distributions which includes, among others, the Weibull, compound Weibull, power function, Gompertz and compound Gompertz distributions. It is assumed that the scale parameter of each component satisfies the inverse power low, the progressive stress is directly proportional to time and the cumulative exposure model for the effect of changing stress holds. Based on type-I censoring, the maximum likelihood estimates (MLEs) of the parameters under consideration are obtained. A special attention is paid to a mixture of two Rayleigh components. Simulation results are carried out to study the precision of the MLEs and to obtain confidence intervals for the parameters involved.     

The experience trap

When companies put seasoned managers in charge of important projects, they don't expect missed deadlines, budget overruns, and rampant defects. However, that's what researchers found when they tested hundreds of experienced project managers with computer games that simulated software development projects. The study, conducted by two professors from Insead and one from Naval Postgraduate School, strongly suggests that veterans in complex environments suffer a breakdown in the learning process. The research reveals three reasons for the breakdowns: Time lags between causes and effects make it difficult to see how they're connected; fallible estimates color the chain of decisions that determine a project's outcome; and a bias toward the initial goals prevents managers from setting revised, more appropriate, targets when project circumstances change. Sticking to an initial low budget goal after a project grew in scope, for instance, led subjects to ignore quality assurance, which led to soaring defect rates--and costs. Companies can take practical steps to fix the learning cycle. They can provide feedback that shows the relationships between important variables in the environment. Such feedback might reveal, say, the 20-day ramp-up that a new quality assurance team needs before becoming fully effective. Tools that apply formal models to calculate such things as the effect of turnover on team productivity also help. Setting goals for behavior, instead of targets for performance, is critical as well. Finally, firms can create project "flight simulators" that mimic actual learning environments but don't let complexity overwhelm trainees. Managers can continue learning only if they get decision support tailored to the challenges they face. Firms would do well to focus more on training people higher up in the organization and stop leaving them to fend for themselves.