14.03.2014 Views

Basic Analysis and Graphing - SAS

Basic Analysis and Graphing - SAS

Basic Analysis and Graphing - SAS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 9 Bootstrapping 251<br />

Perform a Bootstrap <strong>Analysis</strong><br />

Perform a Bootstrap <strong>Analysis</strong><br />

To perform a bootstrap analysis, right-click on a report in a platform report window <strong>and</strong> select Bootstrap.<br />

Specify the number of bootstrap samples <strong>and</strong> choose whether to use fractional weights or to split the<br />

selected column. For more information, see “Bootstrap Window Options” on page 251. Every statistic in the<br />

report is bootstrapped.<br />

The bootstrapping process samples from the original data. An invisible data table is used to store the<br />

sampled data <strong>and</strong> perform the analysis.<br />

Bootstrap Window Options<br />

After you right-click a column in a report <strong>and</strong> select Bootstrap, the Bootstrapping window appears with the<br />

following options:<br />

Number of Bootstrap Samples The number of times that you want to re-sample the data <strong>and</strong> compute<br />

the statistics. A higher number results in more precise estimates of the statistics’ properties.<br />

Fractional Weights Uses fractional weights in the re-sampling process.<br />

If Fractional Weights is not selected, the following observations are true:<br />

– The sampling weights assigned to the original observations are integers.<br />

– The influence of an observation is determined by the relative weights. For example, if an observation<br />

is assigned a weight of 2, then that observation appears twice in the sample for the current bootstrap<br />

iteration.<br />

– The sampling weights can be zero. Observations with a weight of zero are not included in the<br />

estimation process for the current bootstrap iteration.<br />

If Fractional Weights is selected, the following observations are true:<br />

– The sampling weights can be non-integers. The influence of an observation is determined by the<br />

relative weights.<br />

– The sampling weights cannot be zero. All of the observations are included in the bootstrap process.<br />

In both cases, the sum of the weights is n (the number of observations in the original data).<br />

Split Selected Column Splits a stacked column of bootstrap results into separate columns, one for each<br />

statistic. For example, say that the column that you right-clicked before choosing Bootstrap has three<br />

statistics in it. In addition to the main results table with all the statistics stacked, another results table is<br />

created with the results for that column, unstacked into separate columns. For an example, see “Stacked<br />

Results Table” on page 251.<br />

Stacked Results Table<br />

The initial results of a bootstrap analysis appear in a stacked results table (Figure 9.5).

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!