Outlier test and analysis method of degradation data under unequal error variances

For degradation data, it's usually difficult to judge and elimination outlier data, especially when error variances are unequal. So, in the paper, under the assumption that the degradation data obey the same path model form, a degradation outlier test method is presented based on path model when the error variances are unequal. Since the path model gives the trend term of degradation data, its parameters can be considered as the degradation characterization. If the path model parameters are equal for two groups of degradation data, it shows that the two group data have the same degradation trend term. Otherwise, they have different trend terms, and the data of small size can be considered as outlier. For abnormal degradation data, an outlier analysis method is also proposed. It can test the differences in intercepts and slopes between two degradation path models, and determine whether the abnormity is caused by the differences in intercepts, slopes, or both. At the end, an example is given.