[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Surveys in debian



Andrew Suffield wrote:
> I'm rather fond of "How to lie with statistics" as an introduction to
> what it's all about, not least because it teaches extreme cynicism
> about anything that smells like a statistic.

I like Darrell Huff's book too (and recommended it to my students
when I taught), but I feel it's more of a "statistics consumer"
book. Both it and "Tears" are in lots of libraries and fairly
cheap in paperback, so read both if there's time. If one's trying
to run a survey, "Tears" is more useful.  If it's for reading
survey results, get "How to Lie" first.

Even so, I don't agree that it teaches extreme cynicism. Maybe
your previously-mentioned bitterness is colouring you again. I
think it tries to highlight cunning stunts that people do try,
by conspiracy or cock-up (especially newspapers!), rather than
claim everything is suspect. Two sayings I like:
 "Figures won't lie, but liars will figure" - Charles H Grosvenor
 "In god we trust - all others bring data" - W Edwards Deming

Most of your bullets, I agree with. I don't think that surveys
are as bad as you suggest (even self-selecting ones are a
damn sight better than the often-seen perpetual hand-waving)
and I think they'll fail to convince those who *disagree*
rather than all those who don't agree yet. I do agree that the
odds are in favour of the survey sucking, "don't know" is the
most probably outcome and one needs to avoid trying too hard
to prove oneself right.

I think some points you make are either too detailed for a
"mini-HOWTO" format or covered in cited books. I've linked to
this discussion so that those reading around the subject can
see the scepticism which will greet flaky results!

By the way, I've been disappointed that criticism of "User views
on the FDL issue" has been described as an insult by its author
and generally dismissed. http://bignachos.com/?p=10

That survey was pretty well-run for one of its type and it
gave one clear answer out of five, but an "academic defence"
ego where criticism = personal insult will damage future surveys.

-- 
Thanks,

MJR/slef
http://people.debian.org/~mjr/



Reply to: