problem-11.22

problem-11.22  We fit the ANCOVA model first:
> library(MASS)                 # for stepAIC
> kid.weights$BMI = (kid.weights$weight/2.54)/ (kid.weights$height*2.54/100)^2
> res.full = lm(BMI ~ age + gender, kid.weights)
> res.age  = lm(BMI ~ age, kid.weights)
> res.gender  = lm(BMI ~ gender, kid.weights)
> anova(res.full, res.age)
Analysis of Variance Table

Model 1: BMI ~ age + gender
Model 2: BMI ~ age
  Res.Df  RSS  Df Sum of Sq    F Pr(>F)
1    247 9686
2    248 9686  -1     -0.25 0.01   0.94
> anova(res.full, res.gender)
Analysis of Variance Table

Model 1: BMI ~ age + gender
Model 2: BMI ~ gender
  Res.Df  RSS  Df Sum of Sq    F Pr(>F)
1    247 9686
2    248 9722  -1       -37 0.93   0.33
> stepAIC(res)

...
Call:
lm(formula = BMI ~ 1, data = kid.weights)

Coefficients:
(Intercept)
         17
    
We see that neither variable is significant by stepAIC().