Bootstrapping
Simulation
Techniques,
Statistical Analysis Techniques
Like Monte
Carlo methods, the
bootstrap is a computationally intensive method but differs in that it
doesn’t
rely on an assumed data generation process.
Instead it relies on taking samples from an existing
data set. It
involves the following steps:
- Generate a random sample (with replacement) from the
original data set. The
sample should be
the same size as the original data set.
- Calculate the values for the test statistics or
estimators
using this randomly generated sample, and store the results.
- Repeat steps 1 and 2 many times (often thousands of
times).
- Evaluate the performance of the estimator or the
accuracy of
the test statistic based on the set of stored results e.g. find the
mean and
standard error of the stored results.
See
also:
Monte
Carlo Methods
Cross-validation
|
