NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/sci/math/stat/2849

Text File | 1993-01-24 | 3.6 KB | 83 lines

Newsgroups: sci.math.stat
Path: sparky!uunet!mcsun!fuug!anon
From: [e-mail]
Subject: Re: F distribution and relation to Beta
Message-ID: <[e-mail]>
Sender: [e-mail] (The Anon Administrator)
Organization: Anonymous contact service
X-Anonymously-To: sci.math.stat
Date: Sat, 23 Jan 1993 20:41:00 GMT
Lines: 71

>
> [e-mail] (Dino Triantos 575-4386) writes:
>
> >I'm working on a problem in Mood, Graybill and Boes that asks to find the
> >relationship between the F and Beta distributions.  The problem asks:  If
> >X has an F distribution with m and n degrees of freedom, show that
> >
> >                mX/n
> >         W =   -------       has a Beta distribution.
> >              1 + mX/n
>
> >Any suggestions or solutions would be greatly appreciated!
>
> >Dinos Triantos
> >GM/Midsize Automotive Division
> >Vehicle Systems Analysis
> >C-P-C Engr. Center Rm# G241
> >Fax: [phone]
> >E-mail: [e-mail]
>
> Is this a homework problem or will the solution to this
> statistical issue help GM build better cars?  The net has
> informal prohibitions against providing answers to homework
> problems.  The problem you raise is a very standard homework
> assignment in undergraduate statistics classes.  When I did it as
> a homework problem many years ago I remember that it was a bit
> easier to do it as W = 1/(1+(m/n)F).  It won't help you solve it,
> but a useful insight is that W is often known as R^2.  Good luck!
>

    What rock did this respondent crawl out from under of?

    [I've carefully removed his identity from his response.
I'm more concerned with what he represent that who he is.]

    Here we have an individual working in industry who has a statistics
question and some clown decides to disgrace the profession by going beyond
even stereotypical behavior to insult the client (Is this a homework
problem or will the solution to this statistical issue help GM build
better cars?)  and provide some helpful misinformation (When I did it as a
homework problem many years ago I remember that it was a bit easier to do
it as W = 1/(1+(m/n)F).  Put the two together and you've got intellectual
bullying (or incompentence, depending on the reason for the
misinformation) at its worst.  And no one bothers to take the respondent
to task!

    I've never posted to this group because of my anger at the amount of
abuse and misinformation that gets passed out. I've prefered to send my
advice by private e-mail and take part in a real dialogue with my
"clients".  But I've just decided I have no further interest in even
lurking.

    Wouldn't it have been nice if Mr. Triantos had learned that
statisticians were only too happy to help those who were eager and willing
to learn?  Why not merely have told him

    "The only trick to this problem is that there's no trick!  It's
straightforward.  Just find X in terms of W,
                     X = (n/m) (W/(1-W)), and don't forget the Jacobian!
Also, check the region over which the density is positive.  It's trivial
here but it's important, especially in more than one dimension where even
"great men" have screwed up calculating marginal densities by using the
wrong limits of integration."

    Of course, it's too late to tell him now.  If I were he, I'd be
long gone.  As it is, *I'm* gone after this posting.

-------------------------------------------------------------------------
To find out more about the anon service, send mail to [e-mail].
Due to the double-blind system, any replies to this message will be anonymized,
and an anonymous id will be allocated automatically. You have been warned.