Subject: Re: Nielsons (was:Re: Secret Diary sales)
From: hildreth@cg-atla.UUCP (Lon Hildreth)
Date: 1990-11-02, 10:38
Newsgroups: alt.tv.twin-peaks

In article <8402@darkstar.ucsc.edu> unknown@ucscb.UCSC.EDU (The Unknown User) writes:
> >
> >In article <1990Oct29.065308.4866@agate.berkeley.edu> physi-bs@garnet.berkeley.edu (Marc D. Peters) writes:
>> >>Reason #2: There are very, very few 'Nielsen families' out there. Something
>> >>like <.01% of the viewing audience is one. Statistical variations are possible.
> >
> >Well, I don't have the statistical background to go into this, but I
> >do know that seemingly -TINY- samples can give very very very accurate
> >representations as a whole.. Someone else may want to elaborate on this, but
> >I believe they do use a very significantly large sample for what they're trying
> >to measure.

I took a statistics course back in high school (gee, that would have been
14-15 years ago).  My memory may be a little hazy on the specifics but I
recall our teacher talking about the Nielson ratings.  At the time they had
about 1200 sampled households.  I think they've increased their sample size
since then.  Anyway, there are statistical methods (which I've long since
forgotten) to show you the accuracy of your sample.  As I recall, there was
nearly a 100% level of confidence that a sample size of 1200 would be at
least 94% accurate.  I think it there was a 95% level of confidence that
it would be at least 97% or 98% accurate.  It was enough to convince me that 
the ratings were pretty accurate even though they often did not match what I 
watched. 

-- Lon Hildreth ...!{decvax,uunet,samsung}!cg-atla!hildreth Agfa Compugraphic or hildreth@cg-atla.agfa.com Wilmington, MA A Cub fan, but not a Bud man.