Estimation of changepoints using an "S-curve" approximation. Formation of confidence intervals for changepoint locations and magnitudes. Both abrupt and gradual changes can be modeled.
This package does changepoint analysis for both abrupt and gradual change models, by fitting an S-curve using nonlinear least squares. Currently just a single changepoint is analyzed, but this can easily be generalized using the "binary segmentation" method. Code for this will be added in the near future. Both single- and multiple-changepoint models are allowd. In the case of abrupt models, an S-curve is used to approximate a step function. The user specifies a steep slope for this. For gradual models, the slope will be estimated by the algorithm. Output consists of the changepoint location, the pre- and post-mean levels, and for the gradual case, the slope. Standard errors (and covariance matrix) are supplied for all estimated parameters, by calling vcov() on the output.