model.matrixqrrvglm.RdCreates a model matrix. Two types can be
returned: a large one (class "vlm" or one that inherits
from this such as "vglm") or a small one
(such as returned if it were of class "lm").
model.matrixqrrvglm(object, type = c("latvar", "lm", "vlm"), ...)an object of a class "qrrvglm",
i.e., a cqo object.
Type of model (or design) matrix returned.
The first is the default.
The value "latvar" is model matrix mainly comprising
of the latent variable values
(sometimes called the site scores).
The value "lm" is the LM matrix directly
corresponding to the formula argument.
The value "vlm" is the big VLM model matrix given C.
further arguments passed to or from other methods.
This function creates one of several design matrices
from object.
For example, this can be a small LM object or a big VLM object.
When type = "vlm" this function calls fnumat2R()
to construct the big model matrix given C.
That is, the constrained coefficients are assumed known,
so that something like a large Poisson or logistic regression
is set up.
This is because all responses are fitted simultaneously here.
The columns are labelled in the following order and
with the following prefixes:
"A" for the \(A\) matrix (linear in the latent variables),
"D" for the \(D\) matrix (quadratic in the latent variables),
"x1." for the \(B1\) matrix (usually contains
the intercept; see the argument noRRR in
qrrvglm.control).
The design matrix after scaling
for a regression model with the specified formula and data.
By after scaling, it is meant that it matches the output
of coef(qrrvglmObject) rather than the original
scaling of the fitted object.
if (FALSE) { # \dontrun{
set.seed(1); n <- 40; p <- 3; S <- 4; myrank <- 1
mydata <- rcqo(n, p, S, Rank = myrank, es.opt = TRUE, eq.max = TRUE)
(myform <- attr(mydata, "formula"))
mycqo <- cqo(myform, poissonff, data = mydata,
I.tol = TRUE, Rank = myrank, Bestof = 5)
model.matrix(mycqo, type = "latvar")
model.matrix(mycqo, type = "lm")
model.matrix(mycqo, type = "vlm")
} # }