On Mon, 20 Mar 2017 14:26:35 -0700, sms
wrote:

Anytime someone doesn't like the results of a study they try to pick it
apart.

Actually, common practice is to first blame someone and then pick
apart the argument. However, I prefer to undermine the study and let
it collapse under its own weight.

If "pick it apart" is an unacceptable method of debating the merits of
a study, what would you consider to be an acceptable method for this
newsgroup? I could use propaganda, various logical fallacies,
anecdotal evidence, my personal feelings, or perhaps fabricate a
contradictory study. Methinks that "pick it apart" is the same as
breaking down the study into individual claims and seeing how each one
holds together under stress.

Perhaps it would be helpful if I explained how I analyze such studies.
I've done it in this newsgroup at least a dozen times, but have never
really explained how it's done. First, I find the original study.
This is the most difficult part because studies are now hidden behind
pay walls, revised continuously, and "edited for publication" in
different lengths and forms. Once I have the original study, I try to
determine who paid for it. That's because the conclusions and summary
of the study are owned by whomever paid for the study, while the
actual data and calculations are owned by the academics, scientists,
students, and statisticians that ran the study. Often these are
different or even in opposition. I then read the study in as much
detail as I have time available. That's when the differences between
the study and the web page announcing the study become apparent. In
medical studies and surveys, I've seen claims that are quite the
opposite of what the research shows, usually because the claims
support a product or remedy. From this point, my approach varies
depending on what I'm trying to demonstrate, prove, denounce, or
evaluate. Usually, pointing out inconsistencies, gross omissions, and
occasionally math errors is sufficient.

In this case, I have been unable to find the study in either the
original Danish or an English translation. Therefore, I have not read
the original and have had to work with a brief summary from some
unknown report or survey that apparently has been quoted and recycled
extensively. The best I could do is point out that the percentage
cited was meaningless without also disclosing the statistical
population (number of participants in the test). This is hardly "pick
it apart". So, I'll pick at it some more.

One problem with claiming that flashing tail lights reduce accidents
is that there just might not be any correlation between tail lights
and accidents at all. Just because two things correlate (follow the
same trends) does not mean that one causes the other. Some ludicrous
examples:
http://www.tylervigen.com/spurious-correlations
So, how does one prove that flashing tail lights actually cause a
reduction in accidents and that the 30% drop was not a coincidence?
Well, one way is play the record backwards. Instead of giving out
tail lights, find a group that has been using flashing tail lights for
some time and take away their tail lights. If accidents increase,
then there just might be a connection. Perhaps programming the tail
lights so that they flash at different rates under the assumption that
a faster flashing rate is more visible and therefore safer. I could
dream up a few more tests, but basically the idea is to do things that
test for a connection between flashing tail lights and accidents.

The other part of the problem is that it's very easy to demonstrate
that something is unsafe. All that's needed is one accident. However,
it's impossible prove that anything is safe because there will always
be accidents caused by coincidence or disconnected correlations.

Have I "picked apart" your one liner sufficiently?

--
Jeff Liebermann
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558