| model.matrix.multiple {Zelig} | R Documentation |
Use model.matrix.multiple after parse.formula to
create a design matrix for multiple-equation models.
model.matrix.multiple(object, data, shape = "compact", eqn = NULL, ...)
object |
the list of formulas output from parse.formula |
data |
a data frame created with model.frame.multiple |
shape |
a character string specifying the shape of the outputed matrix. Available options are
|
eqn |
a character string or a vector of character strings identifying the equations from which to
construct the design matrix. The defaults to NULL, which only uses the systematic
parameters (for which DepVar = TRUE in the appropriate describe.model function) |
... |
additional arguments passed to model.matrix.default |
A design matrix or array, depending on the options chosen in shape, with appropriate terms
attributes.
Kosuke Imai <kimai@princeton.edu>; Gary King <king@harvard.edu>; Olivia Lau <olau@fas.harvard.edu>; Ferdinand Alimadhi <falimadhi@iq.harvard.edu>
parse.par, parse.formula and the full Zelig manual at
http://gking.harvard.edu/zelig
# Let's say that the name of the model is "bivariate.probit", and
# the corresponding describe function is describe.bivariate.probit(),
# which identifies mu1 and mu2 as systematic components, and an
# ancillary parameter rho, which may be parameterized, but is estimated
# as a scalar by default. Let par be the parameter vector (including
# parameters for rho), formulae a user-specified formula, and mydata
# the user specified data frame.
# Acceptable combinations of parse.par() and model.matrix() are as follows:
## Setting up
## Not run:
data(sanction)
formulae <- cbind(import, export) ~ coop + cost + target
fml <- parse.formula(formulae, model = "bivariate.probit")
D <- model.frame(fml, data = sanction)
terms <- attr(D, "terms")
## Intuitive option
Beta <- parse.par(par, terms, shape = "vector", eqn = c("mu1", "mu2"))
X <- model.matrix(fml, data = D, shape = "stacked", eqn = c("mu1", "mu2")
eta <- X
## Memory-efficient (compact) option (default)
Beta <- parse.par(par, terms, eqn = c("mu1", "mu2"))
X <- model.matrix(fml, data = D, eqn = c("mu1", "mu2"))
eta <- X
## Computationally-efficient (array) option
Beta <- parse.par(par, terms, shape = "vector", eqn = c("mu1", "mu2"))
X <- model.matrix(fml, data = D, shape = "array", eqn = c("mu1", "mu2"))
eta <- apply(X, 3, '
## End(Not run)