this gist contains a small script that generates data and looks for suitable linear regression parameters

linearparams.r

# define model 
y = function(x){ 
  return (2*x + 5 + rnorm(length(x)))
}

# generate data 
X  = rnorm(40)
df = data.frame(X, y(X))
ggplot(df, aes(X,y.X.))+geom_point()

# define costs function 
costs = function(df, a, b){ 
  return( min(sum((df[2] - df[1]*a - b )^2),10000) )
}

# find best value 
axis = seq(0,7,by=0.1)
plotty = expand.grid(a=axis, b=axis)
for(i in 1:length(plotty$a)){
  plotty$error[i] = costs(df, plotty$a[i], plotty$b[i])
}
ggplot( plotty, aes(x=a,y=b)) + geom_tile(aes(fill=error))