comparePred: Compare Predictions

Description

Predicted values are obtained at the specified values of primary for each object. If either object1 or object2 have a grouping structure (i.e. getGroups(object) is not NULL), predicted values are obtained for each group. When both objects determine groups, the group levels must be the same. If other covariates besides primary are used in the prediction model, their group-wise averages (numeric covariates) or most frequent values (categorical covariates) are used to obtain the predicted values. The original observations are also included in the returned object.

Usage

comparePred(object1, object2, primary, minimum, maximum,
    length.out, level, ...)

Value

a data frame with four columns representing, respectively, the values of the primary covariate, the groups (if object does not have a grouping structure, all elements will be 1), the predicted or observed values, and the type of value in the third column: the objects' names are used to classify the predicted values and

original is used for the observed values. The returned object inherits from classes comparePred and augPred.

Arguments

object1,object2: fitted model objects, from which predictions can be extracted using the predict method.
primary: an optional one-sided formula specifying the primary covariate to be used to generate the augmented predictions. By default, if a covariate can be extracted from the data used to generate the objects (using getCovariate), it will be used as primary.
minimum: an optional lower limit for the primary covariate. Defaults to min(primary), after primary is evaluated in the data used in fitting object1.
maximum: an optional upper limit for the primary covariate. Defaults to max(primary), after primary is evaluated in the data used in fitting object1.
length.out: an optional integer with the number of primary covariate values at which to evaluate the predictions. Defaults to 51.
level: an optional integer specifying the desired prediction level. Levels increase from outermost to innermost grouping, with level 0 representing the population (fixed effects) predictions. Only one level can be specified. Defaults to the innermost level.
...: some methods for the generic may require additional arguments.

Author

José Pinheiro and Douglas Bates bates@stat.wisc.edu

Examples

Run this code

fm1 <- lme(distance ~ age * Sex, data = Orthodont, random = ~ age)
fm2 <- update(fm1, distance ~ age)
comparePred(fm1, fm2, length.out = 2)