2

To obtain the sample covariance, does one divide by $n$ or by $n-1$?

I have seen both being used. What is the advantage of using $n-1$? Does it make

the covariance unbiased?

Or is it so that in the formula for the sample correlation the $n-1$ cancels with the $n-1$ used in the formula for the unbiased variance?

user133993
  • 651

1 Answers1

1

  1. If it is a sample, it is divided by (n-1).

  2. Yes. (n-1) makes the estimator unbiased.

  3. You are right. (n-1) in the numerator and in the denominator are canceling each other.

greetings,

calculus.

callculus42
  • 30,550