[Rd] Feature request: allow denominator n in var() and sd()

[Jordan Adamson]

Dear R-devel,
Would R Core consider adding an optional argument to var() and sd() to allow calculation with denominator n?
My motivation comes from teaching statistics through empirical distributions. For the empirical distribution assigning probability 1/n to each observation, the variance is exactly

mean((x - mean(x))^2)

The current var() applies Bessel�s correction. That is useful for unbiased estimation under i.i.d. sampling, but the uncorrected empirical moment is also useful for descriptive statistics and for teaching estimation by substituting the empirical distribution for an unknown distribution. These uses arise even when the observations constitute a sample.
One possible argument would be correction = 1, retaining current behavior by default, with correction = 0 selecting denominator n. The same argument in sd() would preserve its relationship with var().
There is already a precedent in cov.wt(..., method = "ML"). Making the choice available directly in var() and sd() would improve discoverability and avoid requiring a separate function for this basic calculation. Consistent support in cov() would also be desirable.
I found a related R-devel discussion from February 2019: https://stat.ethz.ch/pipermail/r-devel/2019-February/077300.html
The argument name and scope could follow R Core�s preferences.
Thank you for considering this request,

Jordan Adamson
Assistant Professor, Economics Department
Thompson Rivers University


	[[alternative HTML version deleted]]