Periodically correlated time series and the Variable Bandpass Periodic Block Bootstrap

Area of Science:

• Statistics
• Time Series Analysis
• Signal Processing

Background:

• Bootstrapping is a statistical method for estimating sampling distributions.
• Block bootstrapping is used for correlated time series data.
• Existing methods struggle to preserve correlations in periodically correlated time series.

Purpose of the Study:

• Introduce a novel resampling method for periodically correlated time series.
• Address limitations of existing block bootstrapping techniques.
• Demonstrate the advantages of the new method.

Main Methods:

• Developed the Variable Bandpass Periodic Block Bootstrap (VPBB).
• Utilized bandpass filters for frequency separation.
• Applied VPBB to periodically correlated time series data.

Main Results:

• VPBB effectively separates periodically correlated components from noise.
• Simulation studies show significant improvements over prior methods.
• The method preserves correlation structures in time series.

Conclusions:

• The Variable Bandpass Periodic Block Bootstrap offers superior performance.
• This novel approach enhances the analysis of cyclostationary processes.
• VPBB provides a more accurate statistical inference for time series data.