predict.hxlr
Predict/Fitted Values for HXLR Fits
Obtains various types of predictions/fitted values for heteroscedastic extended logistic regression (HXLR) models.
- Keywords
- regression
Usage
# S3 method for hxlr
predict(object, newdata = NULL, type = c("class", "probability",
"cumprob", "location", "scale"), thresholds = object$thresholds,
na.action = na.pass, …)
# S3 method for hxlr
fitted(object, type = c("class", "probability",
"cumprob", "location", "scale"), …)
Arguments
- object
an object of class
"hxlr"
.- newdata
an optional data frame in which to look for variables with which to predict.
- type
type of prediction:
"probability"
returns a data frame with category probabilities,"cumprob"
returns cumulative probabilities,"location"
and"scale"
return the location and scale of the predicted latent distribution respectively, and"class"
returns the category with the highest probability. Default is"class"
.- thresholds
optional thresholds used for defining the thresholds for types
"probability"
,"cumprob"
, and"class"
. Can differ from thresholds used for fitting. If omitted, the same thresholds as for fitting are used.- na.action
A function which indicates what should happen when the data contain
NA
s. Default is na.pass- …
further arguments passed to or from other methods.
Value
For type "prob"
a matrix with number of intervals (= number of
thresholds + 1) columns is produced. Each row corresponds to a row in newdata
and contains the predicted probabilities to fall in the corresponding interval.
For type "cumprob"
a matrix with number of thresholds columns is
produced. Each row corresponds to a row in newdata
and contains the
predicted probabilities to fall below the corresponding threshold.
For types "class"
, "location"
, and "scale"
a vector is
returned respectively with either the most probable categories ("class"
)
or the location ("location"
) or scale (scale
) of the latent
distribution.