Difference between revisions of "Bsplines.R"

From Organic Design wiki
(Seems to be a bug in variable naming from 'fit')
(fixed bs bug)
 
Line 19: Line 19:
  
 
# p41
 
# p41
par(mfrow=c(2,1))
+
par(mfrow=c(3,1))
 
knots <- c(1:3,5,7,8,10)
 
knots <- c(1:3,5,7,8,10)
atx <- seq(0:11, by=0.1)
+
atx <- seq(0, 11, by=0.1)
for(ord in 1:4) {
+
for(ord in 2:4) { # ord cannot be 1
B <- bs(x=atx, degree=ord, knots=knots, intercept=TRUE, Boundary.knots=c(0,11))
+
B <- bs(x=atx, degree=ord-1, knots=knots, intercept=TRUE, Boundary.knots=c(0,11))
 +
        # B <- predict(B, seq(0,11, length=101) # works
 
matplot(atx, B[,1], type="l", ylim=0:1, lty=2)
 
matplot(atx, B[,1], type="l", ylim=0:1, lty=2)
 
matlines(atx, B[,-1], col=1, lty=1)
 
matlines(atx, B[,-1], col=1, lty=1)

Latest revision as of 03:10, 17 October 2007

Code snipits and programs written in R, S or S-PLUS library(splines)

  1. p38

pos <- function(x) ifelse(x>0, x, 0) x <- 1:7 y <- c(8,3,8,5,9,14,11) knot <- 4 plot(x,y)

  1. X <- cbind("int"=1,"x"=x,"x2"=x^2,"x3"=x^3, "xknot"=pos(x-knot)^3)
  2. fit <- lm(y~X-1)

fit <- lm(y~1+x+x^2+x^3+pos(x-knot)^3) xx <- seq(1,7, length=200) XX <- cbind("x"=xx) lines(xx, predict(fit, as.data.frame(XX))) # Bug here abline(v=knot, lty=2) X

  1. p41

par(mfrow=c(3,1)) knots <- c(1:3,5,7,8,10) atx <- seq(0, 11, by=0.1) for(ord in 2:4) { # ord cannot be 1 B <- bs(x=atx, degree=ord-1, knots=knots, intercept=TRUE, Boundary.knots=c(0,11))

       # B <- predict(B, seq(0,11, length=101) # works

matplot(atx, B[,1], type="l", ylim=0:1, lty=2) matlines(atx, B[,-1], col=1, lty=1) title(paste("B splines of order", ord)) abline(v=knots, lty=2) }