After further thought, I no longer think there was an error in my original blog post. My blog post was about estimating a p-value for a Monte Carlo simulation. As shown in this PROC FREQ documentation, the standard error for the Monte Carlo estimate of a binomial proportion has an N-1 in the denominator. It hardly matters when you are doing 10,000 Monte Carlo simulations, but N-1 is the correct denominator. (The OP for this problem should still use N, since that is correct for the SE for the usual estimator p_hat = k/N.
... View more