predict_sf_regression function

Prediction for a scalar-on-function regression model

This function is used to make prediction for a new set of functional and scalar (if any) predictors based upon a fitted scalar-on-function regression model in the output of rob.sf.reg.

predict_sf_regression(object, Xnew, Xnew.scl = NULL)

Arguments

• object: An output object obtained from rob.sf.reg.
• Xnew: A list of matrices consisting of the new observations of functional predictors. The argument Xnew must have the same length and the same structure as the input X of rob.sf.reg.
• Xnew.scl: A matrix consisting of the new observations of scalar predictors. The argument Xnew.scl must have the same length and the same structure as the input X.scl of rob.sf.reg.

Returns

An $n_{test} \times 1$-dimensional matrix of predicted values of the scalar response variable for the given set of new functional and scalar (if any) predictors Xnew and Xnew.scl, respectively. Here, $n_{test}$, the number of rows of the matrix of predicted values, equals to the number of rows of Xnew and and Xnew.scl (if any).

Author(s)

Ufuk Beyaztas and Han Lin Shang

Examples

set.seed(2022)
sim.data <- generate.sf.data(n = 400, n.pred = 5, n.gp = 101, out.p = 0.1)
out.indx <- sim.data$out.indx
indx.test <- sample(c(1:400)[-out.indx], 120)
indx.train <- c(1:400)[-indx.test]
Y <- sim.data$Y
X <- sim.data$X
Y.train <- Y[indx.train,]
Y.test <- Y[indx.test,]
X.train <- X.test <- list()
for(i in 1:5){
  X.train[[i]] <- X[[i]][indx.train,]
  X.test[[i]] <- X[[i]][indx.test,]
}
gp <- rep(list(seq(0, 1, length.out = 101)), 5) # grid points of Xs

model.tau <- rob.sf.reg(Y.train, X.train, emodel = "robust", fmodel = "tau", gp = gp)
pred.tau <- predict_sf_regression(object = model.tau, Xnew = X.test)
round(mean((Y.test - pred.tau)^2), 4)        # 1.868 (tau method)