[ExI] [Bulk] Why so much published 'science' is wrong.
spike
spike66 at att.net
Sat Jul 11 16:16:42 UTC 2015
>... On Behalf Of spike
Subject: Re: [ExI] [Bulk] Why so much published 'science' is wrong.
>... On Behalf Of BillK
...
>>...Statistical significance has nothing to do with actual significance,
though. A statistically significant effect can be trivially small. Or even
completely illusory.
----------
>>...I've got a feeling that this is especially relevant to ESP research,
where much of the claimed effect might probably be just statistical
creations. BillK
_______________________________________________
>...BillK...Even with sharp students at their prime, I fear we would be
appalled at how many draw the wrong conclusions...spike
_______________________________________________
BillK's comment has me thinking about a topic we discussed here a few years
ago. A proposal was made at an education conference (don't know when or
where, would like to know) where the presenter proposed a revamping of
standard engineering education. Currently the standard curriculum requires
four quarters of calculus, then differential equations, and a couple
(sometimes three (and five electives will get you a second major in math))
quarters of more advanced math electives such as multivariate calculus of
variation, complex variables, matrix algebra, all that kind of cool stuff.
In all that, there is only one quarter of statistics required for most
engineering bachelor's degrees, with a second quarter usually offered as an
elective.
Someone at an engineering education conference proposed replacing the
calculus series with a statistics series: make it one quarter of calculus
where you get right to the point, explain what the integral and the
differential functions do and forget teaching the mainstream students all
those now nearly useless integration techniques. Show them how to use
Wolfram's magic trick on the computer, how to do implicit integration and
hit the high points, how to set up a spreadsheet or Matlab routine to do
numerical integration, then don't worry about all those integration
techniques which are never used in the real world but eat up a lot of
classroom time. Then use those three (or four in some cases) quarters to
teach the right ways to use statistics.
I was horrified when I first heard it. Engineering students have been
required to master calculus since about a week after Newton and Leibniz
discovered it. The methods as taught haven't changed much at all in the
last couple hundred years. This would be a major change.
But the idea started growing on me immediately. As I heard it, the
engineering education conference at which it was proposed reacted similarly,
with plenty of the attendees thinking it is a grand idea. I think I have
joined that camp: reduce the calculus, pound on the statistics. The USA and
Britain educate a big fraction of the world's engineers and scientists, so
we really need to get this right. Explain to the students the right way to
use the concept of a null hypothesis. Don't worry about it if they can't
integrate or differentiate, but don't give away any science or engineering
degrees to anyone who doesn't understand the concept of statistical
significance.
Anyone here up to speed on that proposal?
spike
More information about the extropy-chat
mailing list