tsbootstrap: Enhancing Time Series Analysis with Advanced Bootstrapping Techniques

TL;DR

This paper presents tsbootstrap, a Python package that improves time series analysis by providing advanced bootstrapping techniques that account for temporal dependencies, thereby enhancing uncertainty estimation accuracy.

Contribution

It introduces a comprehensive suite of bootstrapping methods tailored for time series data, addressing limitations of traditional approaches.

Findings

Enhanced accuracy in uncertainty estimation for time series.

Seamless integration with Python data science tools.

Supports advanced bootstrapping techniques like Markov and Sieve methods.

Abstract

In time series analysis, traditional bootstrapping methods often fall short due to their assumption of data independence, a condition rarely met in time-dependent data. This paper introduces tsbootstrap, a python package designed specifically to address this challenge. It offers a comprehensive suite of bootstrapping techniques, including Block, Residual, and advanced methods like Markov and Sieve Bootstraps, each tailored to respect the temporal dependencies in time series data. This framework not only enhances the accuracy of uncertainty estimation in time series analysis but also integrates seamlessly with the existing python data science ecosystem, making it an invaluable asset for researchers and practitioners in various fields.