So I have a couple problems I'm working on, for homework.  I was hoping I
could get a little guidance as to where I should go next.
First problem, the one I have the least idea on:
Let x| and s^2 be sample estimates drawn randomly from a population with
mean mu and variance sigma^2, and finite third central moment mu3.
Find Cov(x|,s^2)

So, I set up the problem, trying to get the cov() into a form that I can do
something with. I start with
Cov()=E((x|-mu)(s^2-[(n-1)/n]sigma^2)
and I write out the summation for x| and s^2, since they are calculated from
n samples, and mess around with those(adding a subtracting mu and whatnot)
and manipulate it, and for the expression (x|-mu)(s^2-[(n-1)/n]sigma^2),  I
get

1/n(x|-mu)*Sum[(xi-mu)^2]+n(x|-mu)^3-((n-1)/n)sigma^2 * (x|-mu)

Does that look right? so what am I going to get when I take the expectation
of this? how much simpler can I get it?  it's not immediately apparent to me
what sort of expression I'm supposed to manipulate cov() into, so I could be
on the wrong track.
Second question:
I have a better feel for this, in that I have an Idea of what method I'm
meant to  be using.
Find the pdf, and hence  the expected value and variance, of the estimator
x|.
So, the method I'm meant to use, is use the moment generating function to
find the distribution of Sum[xi], ad then the method of transformation to
find x|=(1/n)*Sum[xi].  So that does sound too bad.  While looking for the
mgf, I want to be sure this step is valid;
mgf(Sum[xi])= E(e^[t*Sum[xi]])=E(product(e^(t*xi)))=Product(E(e^(t*xi)))

Where Product() denotes take the product over i.
Is that last step valid?
And what sort of transformation should I take?

Anyway, thanks for any help you can give.

--
--------------------------------------------------------------------------
Dennis O'Dea
University of Illinois, Department of Economics
www.students.uiuc.edu/~dodea
[EMAIL PROTECTED]


.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to