On Stationarity of Variance Calculation Series

Abstract

While making reliability observations, more samples mean one can make a statistically representative prediction. It is possible to model the failure arrival characteristics statistically using this knowledge. As a natural product of many experiments, a mean and variance figure can be identified for modelling the different occurrences. Even though the different situations can be modelled with such parameters, it may not wholly outline the condition of the product being developed and under test. The variance calculation series derived from the original reliability observation series, which is normally used for simple variance calculation, can be an important consideration. This consideration is rarely encountered. With a mean and a variance figure, a statistical prediction can be made. However, with the very same parameters, another reliability characteristic possessing product or a subcomponent may exist. For this instance, identifying whether the variance calculation series has stationarity and incorporating it in calculations can yield a possible prediction of a more accurate statistical model. In this study, the variance calculation series is considered for their stationary character at hand and is shown to possess such character yielding further modelling possibilities and emphasizing the importance of this consideration.