Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!decwrl!reid
From: reid@decwrl.dec.com (Brian Reid)
Newsgroups: news.config
Subject: Re: new survey to supplement arbitron. Please run this program.
Message-ID: <91@jove.dec.com>
Date: 22 May 89 17:25:06 GMT
References: <80@jove.dec.com> <3215@epimass.EPI.COM> <1735@papaya.bbn.com> <2836@emerald.indetech.uucp> <6796@medusa.cs.purdue.edu>
Distribution: na
Organization: DEC Western Research
Lines: 17

In article <6796@medusa.cs.purdue.edu> spaf@cs.purdue.edu (Gene Spafford) writes:
>
>Keep a month's worth of news?  Ha!  That stopped being possible years
>ago.

We keep 50 days of news online. I need at least a month online to do the
arbitron analysis, and the rest is a safety factor.

>I'm sure that Brian's numbers will be skewed some by this.

No they won't at all. It's easy to tell how much news is there, and the flows
are all percentages anyhow. If you keep 100 days of news online, then the
aggregate flows can be computed to a greater accuracy than if you keep 1 day
of news online, but there's plenty of unskewed information in any amount of
stored-up news. Having 100 sites report on the contents of their 1-day
spooling directory has a lot more information in it, unskewed, than having 1
site report on 100 days.