Fast estimation and choice of confidence interval methods for step regression

In this paper we propose a new fast grid search algorithm for finding the least square estimators of step regression model. This makes it practical to compute resampling-based confidence intervals models. We introduce five data generating models, including one where mean model is (model correctly specified) and four models are not misspecified), use them study coverage probabilities two types regression: symmetric percentile bootstrap subsampling using set rules-of-thumb select block size. Our results show that when specified, Efron provide close-to-nominal have shorter than methods; misspecified, method provides good double bootstrap-like procedure size selection. Finally, apply proposed methods real world environmental dataset on relationship between grassland productivity, soil moisture anomalies other hydro-climatic land variables inference threshold in anomalies, across which there jump productivity.