Related Experiment Videos
On studentising and blocklength selection for the bootstrap on time series.
M Peifer1, B Schelter, B Guschlbauer
1Freiburg Centre for Data Analysis and Modelling, Eckerstr. 1, 79104 Freiburg, Germany. peifer@fdm.uni-freiburg.de
Biometrical Journal. Biometrische Zeitschrift
|August 2, 2005
Summary
We developed a new blockwise bootstrap method for analyzing dependent data like time series. This approach improves statistical testing for medical time series by optimizing block length selection.
Area of Science:
- Statistics
- Time Series Analysis
- Medical Data Analysis
Background:
- Non-parametric bootstrap is standard for independent data.
- Traditional bootstrap fails for dependent data, including time series.
- Existing blockwise bootstrap methods require optimal block length selection.
Purpose of the Study:
- Propose a method for selecting the optimal block length in blockwise bootstrap.
- Enhance the finite sample properties of the blockwise bootstrap.
- Apply the studentised blockwise bootstrap for hypothesis testing on medical time series.
Main Methods:
- Developed a novel method for optimal block length selection in blockwise bootstrap.
- Incorporated studentised statistics to improve bootstrap performance.
- Approximated studentisation for smooth function models efficiently.
- Applied the studentised blockwise bootstrap to medical time series data.
Main Results:
- The proposed method effectively selects optimal block lengths for blockwise bootstrap.
- Studentised statistics improve the finite sample properties of the blockwise bootstrap.
- The method demonstrates successful application in hypothesis testing for medical time series.
Conclusions:
- The studentised blockwise bootstrap offers a robust approach for dependent data analysis.
- This method enhances statistical inference for time series, particularly in medical research.
- Optimal block length selection is crucial for reliable bootstrap results with time series data.