Power Transformation, One-Way ANOVA

Servo Example (pp 791-793)

> servo <- data.frame(time=scan(),

+ location=factor(rep(1:3,each=5)))

1: 4.41 100.65 14.45 47.13 85.21

6: 8.24 81.16 7.35 12.29 1.61

11: 106.19 33.83 78.88 342.81 44.33

16:

Read 15 items

> windows(width=4,height=3,pointsize=10)

> with(servo, boxcox(time, location, lambda=seq(-.4,.6,.01),

+   identify=TRUE, main="Box-Cox Transformation\nservo data"))

Identify confidence limits!

> dev.off()

windows 
      2

(servo01.png here)

> servo.lm <- lm(time ~ ., data=servo)

> windows(width=4,height=3,pointsize=10)

> logrfplot(servo.lm,sub="servo data",pch=21,bg="green3",cex=1.5)

hat(theta) 
 0.7046158

Suggested power = 1 − 0.7046158 = 0.2953842. From the result from Box-Cox transformation, a power of 0 is recommended. That logrithmic transformation is recommended.

> dev.off()

null device 
          1

(servo02.png here)