Software for Changepoints Detection

The paper concerns with freely available software as the possibility to solve the task of detection and statistical evaluation of changes in occurrences (counts, rates, or proportions) of some event, when data is collect over time or in dependence on quantitative attribute. Off-line data analysis is assumed and common principles of change point problems solution are described. Freely available software implementing piecewise linear regression, joinpoint model, segmented regression, changepoints, and adaptive regression splines is selected. The article describes and compares the main features of selected software. Public available data (unemployment rates) are used to compare solutions implemented in selected R packages (earth, changepoint, and segmented) and in the stand-alone Joinpoint Regression Program (JP). The results are not identical partly due to the properties of data and mostly due to different approaches to CP solutions. JP support user to avoid some problems with data collected in time, with difficulties related to assumptions of parametric linear model, or multiple hypotheses testing.