[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Fw: @@@ religion vs. secular poll
hi.
well, the postings on religion vs. secular poll are still coming in, though
mostly now through the backdoor, and after i stopped tallying. but anyway,
i got this message from Don Smith (yeah, ~the~ don smith!) a couple
days (weeks?) ago, and i figured i'd share it before i deleted it for good.
Seems don's been wrapping his brain around this result a little, well, see
for yourself.
---Original Message-----
From: Don A. Smith <dasmith at asm2_mit.edu>
To: jg900 at bright_net <jg900 at bright_net>
Date: Wednesday, December 09, 1998 1:19 PM
Subject: @@@ religion vs. secular poll
>Hi,
>
>I was just perusing the archives, and I noticed:
>
>> 19 people responded and the religious path is more footworn. It
>> seems that 11 of you have trod down that trail, whereas (working out
>> the math here as best i can) 8 respondents have cleared a trail on
>> their own.
>
>I thought I would mention that this distribution of answers is
>completely statistically consistant with a 50-50 split. With a sample
>size of 19, under the null hypothesis that someone is equally likely
>to come to OtR from either path, an 11-8 split (or worse) is about 48%
>likely to happen by chance. Let me put it this way: I discovered OtR
>through a friend who was really big into the CCM scene, so I guess
>that puts me on the religous path. That makes an even 20, with a 12-8
>split. If I flip a coin 20 times, the probability that I will get
>more than 11 heads is about 25%. That's not exceptional. I would
>have to get at least 14 heads (6% chance) before I started saying the
>coin was weighted, and it wouldn't be really convincing until I got 17
>heads (0.1% chance). Or to put it another way around, let's say the
>eight people currently in your sample are the only eight people you
>ever hear that answer from. If you got five more people to tell you
>that they came down the religious path, you would have a 5% chance of
>getting a 17-8 split purely by random chance, which would be
>borderline acceptable for rejecting the null hypothesis of equal
>likelihood.
>
>That is, in order to say one path is "more footworn", you need to know
>that the probability of getting your answer would be extremely
>unlikely if both paths were *equally* footworn. You can then say that
>the data is *not* consistent with equal traffic (at the 5% level, in
>that last example), which therefore means that there probably *is* a
>bias. (95% likely, in that example.) Usually, you don't reject the
>hypothesis unless you can get 95% or higher. With your current data,
>your first sentence really should read: "20 people responded and
>there's a 75% chance that the religious path is more footworn."
>
>Fun with statistics! I hope this was interesting, or at least helpful,
>and not obnoxious.
>
>Yours sincerely,
>--
>Don Smith The Rossi X-Ray Timing Explorer
>dasmith at space_mit.edu http://www.mit.edu/~dasmith/
>
>"You have to have a Death that you're drawn to, because otherwise it's
>like, 'Hey, it's Death, don't get near me!'"
> - Kevin Wade, one of the screenwriters for _Meet Joe Black_
well, no, Don, i didn't find it obnoxious at all. surprising though,
that we'd have to deviate so far from what looks like a fifty fifty split to
even raise an eyebrow. today sure seemed like a good day to post this
to the list. today was a little eyebrow raising, wasn' t it, fellow
secularists?
(hmmm. i musta missed school that day they were teaching Stats.)
jg
np: david bowie __changestwobowie__ {RCA}