Block Bootstrap Algorithm
(Redirected from Moving Block Bootstrap)
Jump to navigation
Jump to search
A Block Bootstrap Algorithm is a statistical resampling algorithm that preserves data dependency structures by resampling contiguous blocks of observations.
- AKA: Moving Block Bootstrap, Block Resampling Algorithm, Dependent Bootstrap Algorithm.
- Context:
- It can (typically) maintain Temporal Dependency in time series data.
- It can (typically) preserve Spatial Correlation in spatial data.
- It can (typically) resample Data Blocks rather than individual observations.
- It can (typically) estimate Standard Errors for dependent data.
- ...
- It can (often) require Block Size Selection through optimization procedures.
- It can (often) balance Dependency Preservation with resampling variability.
- ...
- It can range from being a Fixed Block Bootstrap Algorithm to being a Random Block Bootstrap Algorithm, depending on its block length strategy.
- It can range from being a Non-Overlapping Block Bootstrap Algorithm to being an Overlapping Block Bootstrap Algorithm, depending on its block selection.
- It can range from being a Circular Block Bootstrap Algorithm to being a Stationary Block Bootstrap Algorithm, depending on its boundary handling.
- It can range from being a Single-Level Block Bootstrap Algorithm to being a Hierarchical Block Bootstrap Algorithm, depending on its dependency level.
- ...
- Example(s):
- Counter-Example(s):
- Standard Bootstrap Resampling Algorithms, which assume independent observations.
- Parametric Bootstrap Algorithms, which model dependency structures explicitly.
- Wild Bootstrap Algorithms, which handle heteroscedasticity rather than dependency.
- See: Statistical Resampling Algorithm, Bootstrap Resampling Algorithm, Time Series Analysis Task, Spatial Data Analysis Task, Dependent Data Structure, Jackknife Algorithm, Performance Estimation Algorithm.