Sample R code for bootstrapping residues for regression

y = rnorm(100) + rep(c(0, 1), each = 50)

x = rep(c(1,2), each = 50)

my = mean(y)

res = lm(y~x)$resid

# Log LR test

lr.obs = sum((y-my)^2) - sum(res^2) # Can replace with "F" like statistic

lr.star = c()

for(boot in 1:1000){

y.star = my + sample(res, replace = T)

res.star = lm(y.star~x)$resid

lr.star[boot] = sum((y.star-my)^2) - sum(res.star^2)

}

pval = mean(lr.star>= lr.obs)

This sample code was provided by Dr. Dipen Sangurdekar.  Boostrapping residue of the empirical data is considered to be more reasonable estimation of p-values.