Hello List
 
Thanks to all who responded to my recent request for comments regarding an unbiased estimator of variance given N sample values with associated weights. And a special thanks to Colin Daly who took the time to write out a very nice mathematical demonstration of an unbiased estimator.  I have seen a number of different equations for calculating the variance given a set of sample values and associated weights and have often wondered which is the "best" one. By "best, I mean the estimator which minimizes bias, or better yet, is unbiased.     
To summarize, an unbiased estimate of the variance calculated from a set of samples with associated weights can be obtained through the following equation:
 
s2 = Sum(w_i) / (Sum(w_i) - Sum(w_i * w_i)) * Sum( w_i * (x_i - xbar) * (x_i - xbar))
where xbar = Sum(w_i * x_i) / Sum(w_i),  i = 1,N.
 
If you are using an equation which provides estimates different from those provided by the equation above, then your estimate of variance is biased. :-)
 
Note, that the weights need not sum to 1.0. However, if your weights do sum to 1.0, then the equation above simplifies to:
 
s2 = 1 / (1 - Sum(w_i * w_i)) * Sum( w_i * (x_i - xbar) * (x_i - xbar))
where xbar = Sum(w_i * x_i),  i =1,N.
 
 
 
 
* By using the ai-geostats mailing list you agree to follow its rules 
( see http://www.ai-geostats.org/help_ai-geostats.htm )

* To unsubscribe to ai-geostats, send the following in the subject or in the 
body (plain text format) of an email message to [EMAIL PROTECTED]

Signoff ai-geostats

Reply via email to