Chapter 7, Moving Beyond Linearity - Question 12
In [182]:
p = 100
n = 1000
itr = 5
set.seed(1)
x = matrix(10*rnorm(p*n), ncol=p,nrow=n)
betas = rnorm(p)
beta_0 = rep(0.7,n)
err = rnorm(n)
betas_hat = rnorm(p)
In [183]:
y = beta_0 + x%*%betas + err
In [196]:
err = c()
for(j in 1:itr){
for(i in 1:p){
Y = y-x[,-i]%*%betas_hat[-i]
model = lm(Y~x[,i])
betas_hat[i] = model$coef[2]
}
intercept = model$coef[1]
pred = intercept + x%*%betas_hat
err[j]=mean((pred-y)^2)
}
In [199]:
intercept
In [202]:
plot(err,xlab="Iterations", type="l")
Four iterations is enough for getting a good result
In [203]:
betas_hat
In [204]:
betas
In [205]:
intercept
In [ ]: