Jump to content

Variance of Median


ku

Recommended Posts

[math]X =^{d} R(0,\theta)[/math] and we obtain five independent observations on X: 1.2, 3.7, 2.1, 5.9, and 4.0.

 

The median [math]\hat{M}[/math] is 3.7 and I'm told that [math]var(\hat{M})=\frac{\theta^2}{28}[/math]. How is this obtained? Do I use the formula [math]var(\hat{M})=\frac{1}{4nf(m)^2}[/math]?

Link to comment
Share on other sites

  • 1 year later...
[math]X =^{d} R(0,\theta)[/math] and we obtain five independent observations on X: 1.2, 3.7, 2.1, 5.9, and 4.0.

 

The median [math]\hat{M}[/math] is 3.7 and I'm told that [math]var(\hat{M})=\frac{\theta^2}{28}[/math]. How is this obtained? Do I use the formula [math]var(\hat{M})=\frac{1}{4nf(m)^2}[/math]?

I would like to know the answer too,I think the formula is only for large enough n,which is the asymptotic behavior,what about small sample size?

Link to comment
Share on other sites

  • 2 weeks later...
[math]X =^{d} R(0,\theta)[/math] and we obtain five independent observations on X: 1.2, 3.7, 2.1, 5.9, and 4.0.

 

The median [math]\hat{M}[/math] is 3.7 and I'm told that [math]var(\hat{M})=\frac{\theta^2}{28}[/math]. How is this obtained? Do I use the formula [math]var(\hat{M})=\frac{1}{4nf(m)^2}[/math]?

 

Could someone please explain what [math]X =^{d} R(0,\theta)[/math] means? What is an equals sign superscripted with a d mean? and is R some well-known probability distribution?

Link to comment
Share on other sites

The X =d normally means "X is distributed according to" and the R would be the probability distribution. But, I am unfamilar with what R would mean, too.

 

Hmm, I looked for examples of probability distributions on wikipedia denoted with an R, I found so far:

 

http://en.wikipedia.org/wiki/Rayleigh_distribution

 

http://en.wikipedia.org/wiki/Rice_distribution

 

But I haven't any real idea what these are or are commonly used for

Link to comment
Share on other sites

R(0,theta) is almost certainly a rectangular uniform distribution from 0 to theta.

 

The expectancy of the median is therefore theta/2. The expectancy of the median squared can be determined by considering the probability of picking a number x and having two others above and two below this number. ie a binomial. Then integrating x^2*f(x) between theta and 0 and then fiddling about with a substitution and factorials to convert the integration into a Beta distribution pdf. Then Var(M) = E(M^2)-E(M)^2

 

 

I am unable to decide whether theta is known or not and this would obviously have a significant bearing on the result

Link to comment
Share on other sites

I've just worked through this and get Variance of theta^2/28

 

The important points here are the pdf of the median m ia given by

 

f(x) = 1/theta*(1-x/theta)^2*(x/theta)^2*5!/(2!2!1!)

 

the first part 1/theta is the pdf of the rectangular uniform solution, the (1-x/theta)^2 is the probability of two others above x, the (x/theta)^2 is the probabilty of 2 below x and the 5!/(2!2!11) is the arrangements of x, two above and two below

 

evaluating the integral of x*f(x) between theta and 0 gives E(M) = theta/2 as required

evaluating the integral of x^2*f(x) between theta and 0 gives E(M^2) = 2*(theta^2)/7

 

Variance = theta^2/28 as required

 

More generally for an odd sample size 2n-1 a general formula can be derived by substituting y = x/theta to convert to a Beta style pdf and then by manipulating the integral to equal 1.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.