如何使用stepAIC进行仿真研究

Question

I have to code a simulation study in R. I have X1,...,X15~N(0,1) explanatory variables and Y~N(2+2*X1+0.8*X2-1.2*X15, 1) and I need to simulate n=100 values and repeat that for iter=100 times. Then, for each linear model created I have to find the best sub-model, using stepAIC. I wrote the following code:

set.seed(1234)
sim <- function (sd) {
n <- 100
p <- 15
X <- matrix(rnorm(n * p), n, p)
mu <- 2 + 2*X[,1] + 0.8*X[,2] - 1.2*X[,15]
Y <- matrix(rnorm(100, mu,sd))
sim<-data.frame(Y,X)
r<- lm(Y~., data = sim)
library(MASS)
r0<-lm(Y~1, data=sim)
res<-stepAIC(r0,k=2,direction="forward", scope=list(lower=~1, upper=r))
return(res$coefficients)
}

sim(1)
oo1<- lapply(1:100, sim)

As I am an inexperienced R-user, I think that I'm doing something wrong. The purpose of the study is to find if within the 100 best sub-models (according to stepAIC), there are models that can find the real one (Y=2+2*X1+0.8*X2-1.2*X15+e). In case, I'm doing the wrong things could I get some help/hints so as to implement it correctly?

Answer 1

Here is a working version of your code:

library(MASS)  
set.seed(1234)

sim <- function(sd, n, p) {
  X <- matrix(rnorm(n * p), n, p)
  mu <- 2 + 2*X[,1] + 0.8*X[,2] - 1.2*X[,p]
  Y <- rnorm(n, mean=mu, sd=rep(sd,n))
  df <- data.frame(Y,X)
  r <- lm(Y~., data=df)
  r0 <- lm(Y~1, data=df)
  res <- stepAIC(r0, k=2, direction="forward", 
           scope=list(lower=~1, upper=r), trace=F)
  return(res$coefficients)
}

n <- 100
p <- 15
sim(1, n, p)
oo1 <- lapply(1:100, sim, n, p)