Guide for the statistical analysis of ageing test data – Part 2: Validation of procedures for statistical analysis of censored normally distributed data
Introduction: Procedures for estimating ageing properties are described in specific test procedures, or are covered by the general documents on test procedures for ageing tests with a specific environmental stress (e.g. temperature, radiation, partial discharges). In many cases, a certain property is determined as a function of time at different ageing stresses, and a time to failure based on a chosen end-point criterion is found at each ageing stress. A plot of time to failure versus ageing stress may be used to obtain an estimate of the time to failure for similar specimens exposed to a specified stress, or to obtain an estimate of the value of stress which will cause failure in a specified time. The physical and chemical laws governing the ageing phenomena may often lead to the assumption that a linear relationship exists between the property examined and the ageing time at fixed ageing stresses, or between certain mathematical functions of property and ageing time, e.g. square root or logarithm. The relationships between property values, ageing time, ageing and ageing exposure temperature are determined by mathematical procedures known as analysis of variance, analysis of covariance and regression analysis. In some cases, the property values cannot be determined because of time or property measurement limitations. Such data are referred to as “censored data”. The mathematical procedures are well known and accepted as valid for complete (uncensored) data. This technical report describes
Scope: This part of IEC 60493 provides an account of the work done in designing and validating the statistical procedures for operations on censored groups of normally distributed data. The relationship to similar operations on complete (uncensored) data is examined, and it is shown that “mixed” or wholly uncensored data groups may be analysed in the same way as wholly censored ones. Attention is drawn to the effect of some variation of group size and extent of censoring, as well as the effect of non-uniform data group sizes.
Contents:
- FOREWORD
- INTRODUCTION
- Scope