A New Method for Estimating Inverse Data from Destructive Regular Storage Life Test

摘要

In the destructive regular storage life test of products, there is "inverse" data sometimes, resulting in inaccurate estimates of its reliability index. Based on the theory of isotonic regression and minimum chi-square estimation, we proposed a new method for processing "inverse" data. First, alter the possible “inverse” of original frequency into the frequency meeting sequence constraint by using PAVA algorithm. Then, estimate reliability parameters by means of minimum chi-square estimation. Comparing with traditional methods, we increased the distribution-test of overall failure probability function, the point estimates and confidence intervals of reliability parameters during storage period of products. Finally, example shows that the coefficient of variation obtained by this method are smaller than MLE and the coefficient of variation changes little when sample size changes, reflecting its superiority for estimating small sample data.