Basic Analysis > Bootstrapping > Overview of Bootstrapping
Publication date: 07/30/2020

# Overview of Bootstrapping

Bootstrapping repeatedly resamples the observations that are used in your report to construct an estimate of the distribution of a statistic or statistics. The observations are assumed to be independent.

In the simple bootstrap, the n observations are resampled with replacement to produce a bootstrap sample of size n. Note that some observations might not appear in the bootstrap sample, and others might appear multiple times. The number of times that an observation occurs in the bootstrap sample is called its bootstrap weight. For each bootstrap iteration, the entire analysis that produced the statistic of interest is rerun with these changes:

the bootstrap sample of n observations is the data set

the bootstrap weight is a frequency variable in the analysis platform

This process is repeated to produce a distribution of values for the statistic or statistics of interest.

However, the simple bootstrap can sometimes be inadequate. For example, suppose your data set is small or you have a logistic regression setting where you can encounter separation issues. In such cases, JMP enables you to conduct Bayesian bootstrapping using fractional weights. When fractional weights are used, a fractional weight is associated with each observation. The fractional weights sum to n. The statistic of interest is computed by treating the fractional weights as a frequency variable in the analysis platform. For information about fractional weights, see Fractional Weights and Calculation of Fractional Weights.

To run a bootstrap analysis in a report, right-click in a table column that contains the statistic that you want to bootstrap and select Bootstrap.

Note: Bootstrap is available only from a right-click in a report. It is not a platform command.

JMP provides bootstrapping in most statistical platforms. The observations that comprise the sample are all observations that are used in the calculations for the statistics of interest. If the report uses a frequency column, the observations in that column are treated as if they were repeated the number of times indicated by the Freq variable. If the report uses a Weight variable, Bootstrap treats it as it was treated in the calculations for the report.

Tip: Bootstrap reruns the entire analysis that appears in the platform report from which Bootstrap is invoked. As a result, Bootstrap might run slowly for your selected column because of extraneous analyses in the report. If Bootstrap is running slowly, remove extraneous options from the platform report before running Bootstrap.