3 ms·
1. Kinda, but not quite. ddof=1 will give an unbiased estimate of the variance. It turns out that an unbiased estimate of the standard deviation isn't possible
by im3w1l 8y ago
1. Kinda, but not quite. ddof=1 will give an unbiased estimate of the variance. It turns out that an unbiased estimate of the standard deviation isn't possible without making assumptions about the probability distribution.
2. If you have a finite population and know all the values, ddof=0 will give you an unbiased estimate of the variance. "These are the heights of every person in my class, what is the variance?"
If you have an infinite population and a finite sample and want to estimate the variance of the infinite population, then ddof=1 is correct.
Then there's also the possibility of correcting for the finite size of the population but that's even less important than correcting for the size of the sample.