Re: speed reading

On 2007-05-11, dwight.thieme@xxxxxxxxx <dwight.thieme@xxxxxxxxx> wrote:
On May 9, 10:27 pm, Aaron Denney <wno...@xxxxxxx> wrote:
On 2007-05-10, Jaimie Vandenbergh <jai...@xxxxxxxxxxxxxxxxxxxxx> wrote:

On 9 May 2007 08:48:11 -0700, "dwight.thi...@xxxxxxxxx"
<dwight.thi...@xxxxxxxxx> wrote:

And in fact, I explained in some detail why there is no evidence that
the bell curve hypothesis is incorrect. I'll say it again: "95% of
all adult people are between four and six feet tall" does _not_ imply
the existence of people twelve feet tall.

Well, if that were the actual 95% limits, I'd not be so quick to dismiss
the existence of 12 foot tall people.

In fact, 90% of American men in the early 1970s were between 64.4"
(5'4.4") and 73.6" (6'3.6")[1], which are the 5th and 95th percentile
respectively. Extending by 0.9 inch on either side should easily get to
95% in a width of 11".

The guinness book of world records[2] has an American man that is
8'11", or 38" higher than the mean and median of 69". To find a height
equivalently off the bell-curve for the statistics that Dwight pulled out of
his ass, expand a variance of 11" to 24", which expands the deviation
of 38" to 91". 91" + 5'(60") is 151", or 12'7".

Actually, I just used those figures as an illustration.

As an illustration to make the point that it's ridiculous to consider
anything far outside the 95% range. In your particular case, you were
asserting that 12 feet high if the 95% range is 4-6" is not going to be
observable. My point was that for height, the bell curve is /not/ a
reasonable approximation for the outliers. Things outside the 95% zone
do show appreciable probability of being at large distances.

And, sorry to say this about your 'figuring', but, statistics just
doesn't work that way. I take it you've never had a formal class.
I'll expund on this below.

It works if you assume that height should have a characteristic shape.

Nor, in fact, did I say that it did - I simply said that you cannot
use statistics in this fashion to argue that people like this must
exist.

You argued, rather that a 95% in this range, means that occurences at
the high end, far outside this range, should be vanishingly rare to
non-existant. This doesn't follow at all. I used the height data to
show that for the real distribution of heights, someone as far outside
95% exists as would a 12' high person for the imaginary statistics you
gave would be, when you used such statistics to argue that people
shouldn't be willing to believe a 12' high person given those
statistics. This is just a matter of linear rescaling, not assuming
any particular distribution.

But let's run a few numbers: everyone seems to accept the "95% of all
people read between 200 and 400 wpm" quoted in the one cite.

No, most of us haven't actually accepted that -- we're just saying that
even were that true, it wouldn't show squat.

You also
seem to think that this distribution is a 'bell curve".

No, actually, I don't. A Gaussian distribution is merely the best
distribution to use that assumes as little as possible.

Ok. Then the
median is 300 wpm and sigma is 51 wpm. So, if we look at the
proportion of all people who read from 0 to 600 wpm, then according to
my trusty Ti-83, they comprise 99.99999959% of the population. That
is, only four people out of a billion - yes, that's billion - would be
expected to read faster than 600 wpm.

And there's many reasons to not think that a gaussian is an adequate
summary of the actual data.

Going up to 700 wpm, we get, from purely statistical arguments(and
_your_ assumptions), that only one person in 5,000,000,000 reads that
fast. And all of him posts here ;-)

I never assumed a Gaussian, nor did anyone here.

--
Aaron Denney
-><-
.

Relevant Pages

  • Re: speed reading
    ... the existence of people twelve feet tall. ... equivalently off the bell-curve for the statistics that Dwight pulled out of ... people read between 200 and 400 wpm" quoted in the one cite. ...
    (rec.arts.sf.written)
  • Re: Testing for product of Gaussians
    ... >I have a value created by a computer simulation. ... test against the computed Bessel function distribution will ... I know the answer to the diffusion equation, ... are those of the Statistics Department or of Purdue University. ...
    (sci.math)
  • Re: Most valuable poster
    ... Ooh, the Poisson distribution! ... If you actually *knew* anything about statistics, ... probabilities I want, just like you do. ...
    (talk.origins)
  • Re: Fishers Exact Test and Chi-square
    ... The probability distribution implicitly assumes continuous rather than ... chi-square or normal distribution would only give an approximation ... merely re-iterating that the chi-square statistic for a 2x2 contingency ... there at the least dozes and dozens of chi-square statistics. ...
    (sci.stat.consult)
  • Re: FFT test with few kbits
    ... rules for statistics according to FFT. ... I personally don't care. ... > empirical distribution of a statistic. ... > determine the number of simulations that you need. ...
    (sci.crypt)