Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: prtest and the correction for continuity


From   "FEIVESON, ALAN H. (AL) (JSC-SK) (NASA)" <alan.h.feiveson@nasa.gov>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   st: RE: prtest and the correction for continuity
Date   Fri, 10 Sep 2004 12:08:12 -0500

Richard - If you forget about doing a hypothesis test and just consider the
question of how well you can approximate the cumulative distribution of a
binomal RV by that of a Normal, then you will see that the approximation is
better if you use the continuity correction.

By this argument, one would expect that in general, the test level should be
closer to the nominal level if you use the correction - but this is not a
proof. I would guess for some combinations of p n and k (# of successes)
it's possible that the error induced by not using the contiuity correction
would cancel the bias of the binomial test.

Al Feiveson


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu]On Behalf Of Richard
Williams
Sent: Friday, September 10, 2004 9:06 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: prtest and the correction for continuity


As a sort of followup to my question about binomial confidence intervals 
(you can tell what is coming up in my classes soon) -- In the single sample 
case, -prtest- uses the normal approximation to the binomial, i.e. it 
relies on the fact that, as the number of binomial trials in an experiment 
increases, the distribution of the successes becomes more and more like a 
normal distribution.  However, it isn't really normally distributed, 
because you can't have, say, 12.5 successes.

Some authors therefore recommend a correction for continuity -- if the 
number of successes is more than hypothesized, you subtract .5; if less, 
then you add .5.

-prtest- does not do the correction for continuity nor does it offer an 
option for doing so.  Is this because

(a) the correction has now come to be widely discredited, or

(b) nobody ever asked for it before?

I've seen several sources use the correction and several not use it.  I 
also get the impression that the correction is controversial because it 
sometimes makes matters worse rather than better, but I'm not clear on what 
those conditions are.  Of course, it is best just to use the -bitest- 
command, but if you had to use the normal approximation to the binomial, 
should you apply the correction, or not?  (Put another way, have I been 
misleading my students for 15+ years by teaching the correction for 
continuity?)


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW (personal):    http://www.nd.edu/~rwilliam
WWW (department):    http://www.nd.edu/~soc

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index