Clustering approach for slope change detection in trended data

Most traditional statistical tools have been developed under the assumption of zero slope behavior over time and normally distributed data. However, living systems are systems that tend to have non-normal behavior, exhibit a pattern of constant growth or a steady decrement, and achieve states of dynamic equilibrium, making them hard to manage. This paper evaluates the feasibility of the clustering technique with the median test to estimate slope change points in control charts for data with defined trend with normal and non-normal behavior. A conceptual model is proposed and its implications are discussed.