Narrowest Significance Pursuit: inference for multiple change-points in linear models

We propose Narrowest Significance Pursuit (NSP), a general and flexible methodology for automatically detecting localised regions in data sequences which each must contain change-point (understood as an abrupt change the parameters of underlying linear model), at prescribed global significance level. NSP works with wide range distributional assumptions on errors, guarantees important stochastic bounds directly yield exact desired coverage probabilities, regardless form or number regressors. In contrast to widely studied “post-selection inference” approach, paves way concept “post-inference selection”. An implementation is available R package nsp.