H0: The distribution is Bernoulli
HA: The distribution isn't Bernoulli
In other words, from the way the question is given, it seems like we are
testing the distribution, not the parameter.
-----Original Message-----
From: Paul_Vendetti@aaic.fb.com [SMTP:Paul_Vendetti@aaic.fb.com]
Sent: Friday, February 19, 1999 9:30 AM
To: Dnpowell71@aol.com
Cc: studygroup4B@lists.casact.org
Subject: Re: chi-square goodness of fit
I could be wrong ,but
The question implies that we need the exact value of p (the
parameter is not
estimated) the d.f is n-1, here 2 classes minus 1. You subtract the
extra 1
when have an estimated value of the parameter p.
Dnpowell71@aol.com on 02/18/99 09:26:41 PM
To: studygroup4B@lists.casact.org
cc: (bcc: Paul Vendetti/AAIC)
Subject: chi-square goodness of fit
H E L L O . . . is anybody out there?
11/97 Q20
observe 100 risks, 80 have 0 claims, 20 have 1 claim.
Ho: number of claims per risk follow a bernoulli with mean p
use a chi square goodness of fit test
determine the smallest value of p for which
Ho will be accepted at the .01 significance level.
chi square table: DF 1, 2, 3 and at significance level .01,
corresponding x values are 6.63, 9.21, 11.34
1) how would you do this one?
2) Why is the DF = 1?
I think the DF should be zero, 2 categories minus 1 minus 1 for the
estimated parameter p
2-1-1 = 0 DF. The answer seems to imply that p is for the
population, it is
known.
I think p should be an estimate, otherwise we wouldn't be selecting
p based
on the criteria of accepting Ho at the .01 significancel level in
the first
place.
DP