Xref: utzoo news.admin:7917 news.groups:15481
Path: utzoo!attcan!utgpu!watserv1!watmath!looking!brad
From: brad@looking.on.ca (Brad Templeton)
Newsgroups: news.admin,news.groups
Subject: Re: Fixing the unbroken
Message-ID: <58905@looking.on.ca>
Date: 8 Dec 89 22:51:05 GMT
References: <57577@looking.on.ca> <12074@cbnews.ATT.COM>
Organization: Looking Glass Software Ltd.
Lines: 34
Class: discussion

We come back to the original question.  People have various opinions
about how well the process has worked.  Perhaps it is important, before
proposing any method, to define some sort of metric for the success of
that method.

Admittedly with some schemes the metric *is* the method, as might appear
to be the case for readership measurements.  But with trial groups, it
would be worthwhile comparing readership from the trial period and after
creation (for successes) to see how good a predictor the trial distribution
was.

And yes, the trial distribution could be skewed.  But not nearly as much
as "voting" is skewed!  And those who feel they are not represented are
thus encouraged to send in reports.  If they do it regularly, then not
only do we get their slightly skewing reports for the trial group at hand,
but their non-biased reports for the other trial groups, and trial groups
to come.

I expect most of trial would get at least a 40% propagation -- perhaps
more.  Alt gets as much as 60 to 70%.   Two special sub-hierarchies,
namely trial.dangerous (for groups like sex, drugs etc. that might be
illegal in some jurisdictions) and trial.commercial (for groups like
misc.forsale, comp.newprod, news.announce.conferences, misc.jobs.* etc.)
would get lower propagation, but not a lot lower, I think.

Pretty good results, I predict.

Unfortunately, we can hardly find out if people want trial groups over
"voting" by holding a "vote!"  Such a result would be as meaningful as,
"amazingly 100% of survey respondents said that they liked responding to
surveys!"

-- 
Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473