J
John Larkin
Guest
On Sun, 22 Oct 06 12:09:48 GMT, jmfbahciv@aol.com wrote:
where one of my guys decided that points on one end of the scale
should be weighted more than points on the other end. He came up with
a page of equations, full of matrices and things, to apply a weighting
function on top of a polynomial regression. It hurt my head, so I
suggested he just copy various multiples of different points back into
the table, like 5x from the low end tapering up to 25x on the high
end, the run the curve fit. You could do that with punch cards, too.
But the problem turned out to be a lot simpler, almost linear, when we
looked at it from another direction. And when we changed the specs on
the product, it got even simpler.
John
We were just talking about doing a polynomial curve fit to a dataset,In article <Qmu_g.14851$GR.13390@newssvr29.news.prodigy.net>,
lucasea@sbcglobal.net> wrote:
jmfbahciv@aol.com> wrote in message
news:ehd5rn$8qk_009@s884.apx1.sbo.ma.dialup.rcn.com...
Common sense would deemd that interval wider than the data.
....and because you have absolutely no background in or understanding of
statistics, your "common sense" would be wrong. Please do learn whereof you
speak, before you speak. The gaps in your knowledge in areas that are key
to a lot of the points you insist are right, is phenomenal and appalling.
I don't know it's wrong. I do know enough that bad data will
never show any statistical significance.
Don't you think *they* are in a better position to judge the quality of
their data than *you* are,
Nope. Not when it's for the BBC comsumption.
since your understanding of statistics is
essentially non-existent? And don't you think that the peers who reviewed
the article and allowed it to be published might also be just a tiny tad
more knowledgeable of statistics than you are?
This is not a question of ability of applying statistics. It is
a question of the agility of applying statistics. I am sceptic
of the agility.
Yes, you do not know enough. Have you studied statistics, sampling, data
analysis?
Yes. A long time ago.
Then you've clearly forgotten everything you learned.
I can certainly open my stat books and yak a good game of
presenting the same data point as a dozen. In the olden
days, I'd just dup 12 cards.
where one of my guys decided that points on one end of the scale
should be weighted more than points on the other end. He came up with
a page of equations, full of matrices and things, to apply a weighting
function on top of a polynomial regression. It hurt my head, so I
suggested he just copy various multiples of different points back into
the table, like 5x from the low end tapering up to 25x on the high
end, the run the curve fit. You could do that with punch cards, too.
But the problem turned out to be a lot simpler, almost linear, when we
looked at it from another direction. And when we changed the specs on
the product, it got even simpler.
John