NSA knows who you've called.

[dan at geer.org]

Nick Owen writes:
-+---------------
 | ...
 | Or to teach pollsters to ask the correct questions.
 | ...


All,

Mr. Owen is dead-on.  Speaking as someone who has had
a formal education in statistics including the design
of survey instruments, I will say that of all the ways
in which it is possible for the dishonest to skew the
results of quantitative analysis, survey design is hands
down the most vulnerable.  You want the numbers to come
out your way?  Sure, you can manipulate any data set of
numbers to lean the direction you want them to lean,
but if you control the survey instrument used to collect
the raw data in the first place you 0wn the analysis
in ways that re-analysis by others cannot erase.

Case in point: Allowing those who care about Issue XYZ
to self-select whether to take your survey guarantees
overweighting the tails of your distribution and in
ways that you may not be able to see (such as organized
survey takers who talk to each other).  Sort of like
an Internet-mailing-list, no?

--dan


---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at metzdowd.com