Distribution Analysis: Distributional Modeling


Estimators Tab

You can use the Estimators tab to specify parametric distributions to fit to the data. (See FigureĀ 15.4.) The options for the Estimators tab correspond to options for the HISTOGRAM statement in the UNIVARIATE procedure. See the documentation in the Base SAS Procedures Guide for details.

For each distribution, you can enter values for one or more parameters, and estimate the remaining parameters with maximum likelihood estimation (MLE). The analysis typically creates a histogram overlaid with density curves, one for each specified distribution.

To add a new distribution to the Estimators list, click a blank item and select a distribution from the Family list.

To delete a distribution from the Estimators list, click an existing distribution and select None from the Family list.

To change a distribution in the Estimators list, click the distribution and select a new distribution from the Family list.

Threshold parameters are set to zero unless you clear the Use threshold of 0 check box, in which case the threshold parameter is estimated by MLE. Other parameters in a distribution are estimated from the data by using MLE, unless you select Manual parameter selection.

The Estimator tab contains the following UI controls:

Estimators


displays a list of distributions that are fitted to the data. Clicking an item in this list enables you to change the distribution or to specify parameters for the distribution. You can specify up to five distributions.

Family


specifies the distribution for the selected item in the Estimators list.

Parameter Selection


specifies how to determine parameters of the selected distribution in the Estimators list. If Automatic is selected, then parameters are estimated by using MLE. If Manual is selected, then you can enter one or more known parameters. Unspecified parameters are estimated by using MLE.

Use threshold of 0


specifies whether the threshold parameter is set to zero for the current distribution. If you clear this check box, then the threshold parameter is estimated by using MLE.

Note: Maximum likelihood estimation of two parameters does not always converge. Three-parameter estimation often does not converge. Three-parameter estimation is attempted if you clear the Use threshold of 0 check box while Automatic is selected.