hw5

Stat 210B: Homework Assignment 5

Due: Wednesday March 12

I. The data cpus contains a relative performance measure and characteristics of 200 CPUs. The variables are

     syct: cycle time in nanoseconds
     mmin: minimum main memory in kilobytes
     mmax: maximum main memory in kilobytes
    cach: cache size in kilobytes
    chmin: minimum number of channels
     chmax: maximum number of channels
     perf: published performance on a benchmark mix relative to an IBM 370/158-3

Build a regression tree to predict the performance by using log(perf) as the response and the other variables as the predictors.

Show your final tree model and explain carefully how you reach it.
How many splits in your final tree? What is the residual sum of squares?
Predict the performance of the following cpu:

syct=23, mmin=16000, mmax=32000, cach=64, chmin=16, chmax=32

II. The data fgl has 138 rows and 10 columns. It was on fragments of glass collected in forensic work. The variables are

    RI: refractive index
    Na: sodium
    Mg: manganese
    Al: aluminum
    Si: silicon
    K: potassium
    Ca: calcium
    Ba: barium
    Fe: iron
    type: WinF, Veh, Con, Tabl or Head.

Build a class tree to predict the type of glass.

Show your final tree model and explain carefully how you reach it.
How many splits in your final tree? What is the misclassification error rate?
Predict the type of the following glass:

RI=-1.69, Na=13.34, Mg=3.57, Al=1.57, Si=72.87, K=0.61, Ca=7.89, Ba=0, Fe=0

Hint: Your tree model should have at least one terminal node for each type.

Note: To grow a larger tree, you can use option "cp" and/or "minsplit" in "rpart". See help(rpart.control).