derivative.scam: Derivative of the univariate smooth model terms

Description

Function to get derivatives of the smooth model terms (currently only of the univariate smooths). Analytical derivatives for SCOP-splines, finite difference approximation is used for all others

Usage

derivative.scam(object,smooth.number=1,deriv=1)

Arguments

object

fitted scam object

smooth.number

ordered number of the smooth model term (1,2,...), ordered as in the formula, which derivative is needed to be calculated.

deriv

either 1 if the 1st derivative is required, or 2 if the 2nd

Value

dvalues of the derivative of the smooth term.
se.dstandard errors of the derivative.

References

Pya, N. and Wood, S.N. (2014) Shape constrained additive models. Statistics and Computing, DOI 10.1007/s11222-013-9448-7

Examples

Run this code

set.seed(2)
n <- 200
x1 <- runif(n)*4-1;
f1 <- exp(4*x1)/(1+exp(4*x1)) # monotone increasing smooth
x2 <- sort(runif(n)*3-1)      # decreasing smooth
f2 <- exp(-1.3*x2)
f <- f1+ f2 
y <- f+ rnorm(n)*0.2
## fit model, results, and plot...
b <- scam(y~ s(x1,k=20,bs="mpi")+s(x2,k=15,bs="mpd"),family=gaussian)

d1 <- derivative.scam(b,smooth.number=1,deriv=1)

par(mfrow=c(1,2))

xx <- sort(x1,index=TRUE)
plot(xx$x,d1$d[xx$ix],type="l",xlab=expression(x[1]),
     ylab=expression(df[1]/dx[1]))

d2 <- derivative.scam(b,smooth.number=2,deriv=1)

xx <- sort(x2,index=TRUE)
plot(xx$x,d2$d[xx$ix],type="l",xlab=expression(x[2]),
     ylab=expression(df[2]/dx[2]))