FREG (version 1.1)

predict.olfreg: Predict OLFREG model

Description

Prediction of OLFREG model

Usage

# S3 method for olfreg
predict(object, ..., newdata = NULL, type = c("probabilities", "labels"))

Arguments

object

OLFREG model for which predictions are computed

...

additional arguments relevant for the generic method

newdata

an optional argument. Newdata should be organized as a list. The elements of the list are covariates from OLFREG model, respectively. No data transformation is needed. Thus, functional covariates are entered in the list newdata in their raw form. The predict.olfreg function will take care of the transformation of such covariates into the functional form of their equivalents from OLFREG model.

type

c("probabilities", "labels")

Value

predictions of dependent variable y

Examples

# NOT RUN {
# cycling dataset
library(fda)
# creation of ordinal variable from HR variable
zoneHR=rep(0,216)
zoneHR[which(rowMeans(cycling$HR[,1:1700])<107)]=1
zoneHR[which((rowMeans(cycling$HR[,1:1700])<125)&(rowMeans(cycling$HR[,1:1700])>107))]=2
zoneHR[which((rowMeans(cycling$HR[,1:1700])<142)&(rowMeans(cycling$HR[,1:1700])>125))]=3
zoneHR[which((rowMeans(cycling$HR[,1:1700])<160)&(rowMeans(cycling$HR[,1:1700])>142))]=4
zoneHR[which((rowMeans(cycling$HR[,1:1700])>160))]=5
# first functional variable - power (WATTS)
watts = t(cycling$WATTS[,1:1700])
# set up a fourier basis system due to its cycling pattern
xbasis = create.fourier.basis(c(1,1700),50) # 50 basis functions for example
watts.fd = smooth.basis(c(1:1700),watts,xbasis)$fd
zoneHR = as.factor(zoneHR)
# additional functional variable - cadence (CAD)
cad = t(cycling$CAD[,1:1700])
# set up a functional variable for cad
xbasis2 = create.bspline.basis(c(1,1700), nbasis = 25, norder = 4)
cad.fd = smooth.basis(c(1:1700),cad,xbasis2)$fd
formula = zoneHR ~ watts.fd + cad.fd
olfreg.model = olfreg(formula = formula)

# Predict with new data included
watts_new = t(cycling$WATTS[,101:1800])
cad_new = t(cycling$CAD[,101:1800])
newdata = list(watts_new, cad_new) # could also be fd var instead of raw data
yhat = predict(olfreg.model, newdata = newdata, type = "labels")
# }