Evaluation of neural net surface.

Usage

predict.nnreg(out, x, model=NA, derivative=0, type="full")

Arguments

out Fitted nnreg object.
x Matrix of x values on which to evaluate the neural net surface.
model Model number to use in predicting. Default is the best model based on GCV(2).
derivative Derivative of function is returned if derivative=1.
type Form of predictions. Default is the prediction for the independent variable. If type="terms" the individual values for the hidden units are calculated.

Value

Vector of predicted responses. If derivative=1 a vector of derivatives or a matrix of partial derivatives is returned. If type="terms" a list with components: u a matrix with the projections of the independent vectors plus the offset ( X%*% gamma + gamma_0) for each hidden unit, yhat, a matrix where the columns vectors are the evaluation of each hidden unit and constant, the value of the constant (intercept) in the model.

See Also

nnreg, predict.surface

Examples

nnreg(ozone$x,ozone$y,1,2) -> fit           # nnreg fit
cbind(seq(87,89,,10),seq(40,42,,10)) -> x   # new x matrix
predict(fit,x) -> out                       # evaluate fit at x

[Package Contents]