Previous: Interchanging the Three Methods

Up: Advanced Zelig Uses

Next: Making the Model Compatible

Adding Models and Methods to Zelig

Zelig is highly modular. You can add methods to Zelig and, if you wish, release your programs as a stand-alone package. By making your package compatible with Zelig, you will advertise your package and help it achieve a widespread distribution.

This chapter assumes that your model is written as a function that takes a user-defined formula and data set (see Chapter ), and returns a list of output that includes (at the very least) the estimated parameters and terms that describe the data used to fit the model. You should choose a class (either S3 or S4 class) for this list of output, and provide appropriate methods for generic functions such as summary(), print(), coef() and vcov().

To add new models to Zelig, you need to provide six R functions, illustrated in Figure . Let mymodel be a new model with class "myclass".

Figure: Six functions (solid boxes) to implement a new Zelig model

139#139

These functions are as follows:

1. zelig2mymodel() translates zelig() arguments into the arguments for mymodel().
2. mymodel() estimates your statistical procedure.
3. param.myclass() simulates parameters for your model. Alternatively, if your model's parameters consist of one vector with a correspondingly observed variance-covariance matrix, you may write two simple functions to substitute for param.myclass():
   1. coef.myclass() to extract the coefficients from your model output, and
   2. vcov.myclass() to extract the variance-covariance matrix from your model.
4. qi.myclass() calculates expected values, simulates predicted values, and generates other quantities of interest for your model (applicable only to models that take explanatory variables).
5. plot.zelig.mymodel() to plot the simulated quantities of interest from your model.
6. A reference manual page to document the model. (See Section )
7. A function (describe.mymodel()) describing the inputs to your model, for use with a graphical user interface. (See Section ).
8. An optional demo script mymodel.R which contains commented code for the models contained in the example section of your reference manual page.