Sorry, I should have put it in my original reply. If you search for where it asks for an implementation of the standard deviation, you'll see that it asks for an implementation including Bessel's correction (using N-1 instead of N for the normalization, used to correct for the one degree of freedom lost by the calculation of the mean). However, when you look at the answer, provided, you'll notice that it divides by N, and does not provide 'NaN' for the condition N=1, as was asked.