Previous: Managing Statistical Model Inputs

Up: Managing Statistical Model Inputs

Next: Single Response Variable Models:

Describe the Statistical Model

After setting up the 53#53 matrix, the next step for most models will be to identify the corresponding vector of parameters. For a single response variable model with no ancillary parameters, the standard R interface is quite convenient: given 53#53 , the model's parameters are simply 12#12 .

There are very few models, however, that fall into this category. Even Normal regression, for example, has two sets of parameters 12#12 and 13#13 . In order to make the R formula format more flexible, Zelig has an additional set of tools that lets you describe the inputs to your model (for multiple sets of parameters).

After you have written down the statistical model, identify the parameters in your model. With these parameters in mind, the first step is to write a describe.*() function for your model. If your model is called mymodel, then the describe.mymodel() function takes no arguments and returns a list with the following information:

• category: a character string that describes the dependent variable. See Section  for the current list of available categories.
• parameters: a list containing parameter sets used in your model. For each parameter (e.g., theta), you need to provide the following information:
  • equations: an integer number of equations for the parameter. For parameters that can take, for example, two to four equations, use c(2, 4).

  • tagsAllowed: a logical value (TRUE/FALSE) specifying whether a given parameter allows constraints.

  • depVar: a logical value (TRUE/FALSE) specifying whether a parameter requires a corresponding dependent variable.

  • expVar: a logical value (TRUE/FALSE) specifying whether a parameter allows explanatory variables.

(See Section  for examples and additional arguments output by describe.mymodel().)