You may not be aware of
the statistical techniques or
the metanalysis used to
compile the
statisticsFYI HK - I
am perfectly well aware of
stats and meta analyses; in
point of fact I'm in the
middle of preparing a meta
analysis of bivalve feeding
rates for publication as we
speak - LOL I just presented
my preliminary results in
seminar this morning :)
but not having read the
journals and seen the
reasearch you cannot address
their validity.This is
my point exactly - I have NOT
read the journals and can NOT
address their validity. I do
not know if you have ever
actually noticed this (I know
that I had no clue until I
actually started to extract
data from papers to do meta
anlyses) but the number of
papers published in
'reputable' and 'elite'
journals that do not report
sample sizes, variances, etc.
is APALLING. It is
unfortunate, but in a majority
of cases the writers,
reviewers, and readers alike
do not actually understand how
various stats work and simply
run them straight from the
textbook. Simply doing an
ANOVA and coming up with a
significant result does not
make your study valid. I need
to see how the data were
collected, what type of data
they were and how they were
analyzed before I can assess
validity.
Since I have not seen the
primary literature, I prefer
to tend to err on the
conservative side and be
skeptical until I have reason
to be otherwise, and hope that
most other people will be
aware enough to do the same.
I wish I could tell you
otherwise, but the stats are
scientifcally and empiracally,
gathered, compiled and
reported. (I can't verify
this in terms of the stuff I
found on the net, but since
they agree with the genuine
stats, I included them
above.)I wish you
could tell me otherwise, too!!
But like I said, I am not
going to trust numbers -
especially ones that seem to
vary so widely even just in
second or thirdhand summary on
the net (I came across numbers
as high as 1 in 6, just
flipping through a few
websites) without more
concrete evidence. I most
certainly am not going to
believe that the numbers
you've quoted are valid simply
because you say that they are
- especially since they seem
to be the most 'outrageous'
ones of all the ones that seem
to be thrown about. I am
probably not really in th e
best position to judge (but at
the same time I know more than
the general populace about
this) but I am extraordinarily
skeptical of studies published
in medical journals. Most of
the people who do these
studies (and I say do, not
design, because most of the
time it seems fairly evident
that there was not a large
amount of forethought given to
the specifics of a given
problem) and most of the
people that review them for
publication have never learned
how to properly design an
experiment, and have most
likely never had anything
beyond a general stats course
in graduate school. Sample
sizes are ridiculously small,
there is rarely replication
(these two are through no
fault of the experimenters;
study on humans is necessarily
difficult in this sense) and
as I've said about ten million
times in this post, stats are
often poorly applied.
Okay, I'm starting to sound
condescending and I don't mean
to; it's just that
experimental design (good
experimental design) and
analysis is a pretty important
topic to me. I know plenty of
people - grad students, post
docs, professors - who you
would hope would understand
these things, but really
don't. Ignorance is bliss, I
suppose :)
<img src="http://gvac.50megs.com/images/toocute2.jpg">
This message was edited by TooCute on 10-31-01 @ 11:58 PM