Previous: Writing New Models

Up: Writing New Models

Next: Describe the Statistical Model

Managing Statistical Model Inputs

Most statistical models require a matrix of explanatory variables and a matrix of dependent variables. Rather than have users create matrices themselves, R has a convenient user interface to create matrices of response and explanatory variables on the fly. Users simply specify a formula in the form of dependent ~ explanatory variables, and developers use the following functions to transform the formula into the appropriate matrices. Let mydata be a data frame.

> formula <- y ~ x1 + x2                   # User input

# Given the formula above, programmers can use the following standard commands
> D <- model.frame(formula, data = mydata) # Subset & listwise deletion
> X <- model.matrix(formula, data = D)     # Creates X matrix
> Y <- model.response(D)                   # Creates Y matrix

where

• D is a subset of mydata that contains only the variables specified in the formula (y, x1, and x2) with listwise deletion performed on the subset data frame;
• X is a matrix that contains a column of 1's, and the explanatory variables x1 and x2 from D; and
• Y is a matrix containing the dependent variable(s) from D.

Depending on the model, Y may be a column vector, matrix, or other data structure.