### *** Motorcycle Data *** ###
library(tree)

library(randomForest)

library(MASS) # data

data(mcycle)

par(mfrow=c(1,2))

mct <- tree(accel ~ times, data=mcycle)

plot(mct, col=8)

text(mct, cex=.75, font=2)

## create a grid of times to predict over

timegrid <- data.frame(times=seq(1,60,length=1000))

## plot my fitted tree surface

plot(mcycle, pch=21, bg=8)

lines(timegrid$times, predict(mct, newdata=timegrid), col=2, lwd=3)

## now do it with random forests

rfmc <- randomForest(accel ~ times, data=mcycle, 
	## works fine without below, but in this example this 
	## gives a better looking tree.  nodesize is min leaf size
	nodesize=10)
	
	
pred.rfmc <- predict(rfmc, timegrid)

plot(mcycle, pch=21, bg=8)

lines(timegrid$times, pred.rfmc, col=2, lwd=2)

## plot doesn't produce a nice dendrogram... why?

plot(rfmc) # this is just the MSE as you increase your sample of trees