Re: Measure of representativeness
- From: "Ray Koopman" <koopman@xxxxxx>
- Date: 31 Jan 2007 10:13:47 -0800
diodoo wrote:
Suppose I have
State | Sample | Population
s1 | n1 | N1
s2 | n2 | N2
. | .. | ..
Is the correlation between (n1 n2 ..) and (N1 N2 ..) a good way to measure the representativeness of the sample? Or are there other standard way to quantify this?
Thanks you,
diodoo
One good way is the uncentered correlation of the square roots
of the counts:
sum sqrt(ni*Ni)
r = -----------------------,
sqrt[(sum ni)*(sum Ni)]
which is related to the "Hellinger distance" between the two
distributions.
.
- References:
- Measure of representativeness
- From: diodoo
- Measure of representativeness
- Prev by Date: Re: How to solve this random integer generation problem?
- Next by Date: Re: questions on how to create linear spline terms
- Previous by thread: Measure of representativeness
- Next by thread: Conditional probability density function with GARCH(1,1) for maximum likelihood estimation
- Index(es):
