R
relic
Twisted said:You should know that Internet polls are inherently unreliable and
unscientific. That is to say, they prove absolutely nothing. Nada.
Zip. Zero. Zilch. Responders are self-selecting -- there's bias
source #1. Anyone might lie. There's source #2. Unless a lot of
hashing and strong crypto is used, errors will creep in from network
unreliability, Microsoft software, and other sources. This won't
produce a systematic bias, but it will make any results approximate,
this destroying any usefulness for making absolute claims rather than
"approximately half of our respondents chose Pepto-Bismol over the
other leading brand" or whatever. So "Everybody except <name>
believes X" and the like are right out. Also, without good crypto,
anyone can stuff the ballot box, alter someone else's vote, or
otherwise corrupt the process, which is bias source #3. There are
probably more bias sources I'm missing, too.
I found her poll to follow the accepted norms of Scientific polling. I
doubted her too, but was satisfied after I looked at her techniques and
data.