The Trend Reliability Growth application uses JMP’s Reliability Growth platform to let users look at changes in beta and lambda parameters over time. It does this by fitting multiple models to subsets of contiguous observations until the entire data set is covered. When the sets of observations are contiguous and non-overlapping, the results are analogous to fitting a Reinitialized Weibull NHPP with a Phase column where each group, except possibly the last, has an equal number of observations. Unlike Reliability Growth with a Phase column, overlapping groups can be fit.
The application assumes data is in time to event format. Window controls the number of observations used for each model. Step corresponds to the step size for determining the starting observations in each subset. Specifically, for each iteration, Window observation are used to fit a Crow-AMSAA model starting at row (i-1)*Step, where i is the iteration number. Subsets will overlap if Step is smaller than Window. If Fixed Starting Point is selected, the first subset starts with observation 1 with Window determining the number of observations. On each subsequent iteration, the window size is increased by Step with the starting point remaining observation 1.
