Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: covariance
Replies: 1   Last Post: Mar 7, 2010 10:42 AM

 Messages: [ Previous | Next ]
 Jack Tomsky Posts: 1,835 Registered: 12/18/04
Re: covariance
Posted: Mar 7, 2010 10:42 AM

> In statistics the sample variance is different from
> population
> variance, is it the same for sample covariance? Do we
> divide by n-1

Let NUM = Sum(Xik-XBARi)(Xjk-XBARj) be the numerator, where the sum is over k = 1, ..., n.

1). The unbiased estimate of the covariance, SIGij, is NUM/(n-1).

2). For samples from a multivariate normal distribution, the maximum likelihood estimate of SIGij is NUM/n.

3). For samples from a multivariate normal distribution, the minimum mean-square estimate of SIGij is NUM/(n+1).

Jack
www.tomskystatistics.com

Date Subject Author
3/7/10 gaya.patel@gmail.com
3/7/10 Jack Tomsky