Define Custom Distributions Using the Distribution Fitter App

You can define a probability object for a custom distribution and use the Distribution Fitter app or fitdist to fit distributions not supported by Statistics and Machine Learning Toolbox™. You can also use a custom probability object as an input argument of probability object functions, such as pdf, cdf, icdf, and random, to evaluate the distribution, generate random numbers, and so on.

Open the Distribution Fitter App

• MATLAB^® Toolstrip: On the Apps tab, under Math, Statistics and Optimization, click the app icon.

• MATLAB command prompt: Enter distributionFitter.

  distributionFitter

  

Define Custom Distribution

To define a custom distribution using the app, select File > Define Custom Distributions. A file template opens in the MATLAB Editor. You then edit this file so that it creates a probability object for the distribution you want.

The template includes sample code that defines a probability object for the Laplace distribution. Follow the instructions in the template to define your own custom distribution.

To save your custom probability object, create a directory named +prob on your path. Save the file in this directory using a name that matches your distribution name. For example, save the template as LaplaceDistribution.m, and then import the custom distribution as described in the next section.

Import Custom Distribution

To import a custom distribution using the app, select File > Import Custom Distributions. The Imported Distributions dialog box opens, in which you can select the file that defines the distribution. For example, if you create the file LaplaceDistribution.m, as described in the preceding section, the list in the dialog box includes Laplace followed by an asterisk, indicating the file is new or modified and available for fitting.

Alternatively, you can use the makedist function to reset the list of distributions so that you do not need to select File > Import Custom Distributions in the app.

makedist -reset

This command resets the list of distributions by searching the path for files contained in a namespace named prob and implementing classes derived from ProbabilityDistribution. If you open the app after resetting the list, the distribution list in the app includes the custom distribution that you defined.

Once you import a custom distribution using the Distribution Fitter app or reset the list by using makedist, you can use the custom distribution in the app and in the Command Window. The Distribution field of the New Fit dialog box, available from the Distribution Fitter app, contains the new custom distribution. In the Command Window, you can create the custom probability distribution object by using makedist and fit a data set to the custom distribution by using fitdist. Then, you can use probability object functions, such as pdf, cdf, icdf, and random, to evaluate the distribution, generate random numbers, and so on.