"Larry C. Lyons" <LarryCLyons@someoneElse.Invalid> wrote in message

Knowing a bit about meta-analysis (see
http://www.lyonsmorris.com/MetaA/index.htm),
I thought I'd respond to this one.

Finally... much appreciated.
It was getting frustrating... as usual,
I was contemplating making egregious errors
just to get more participation...
well maybe I did, yeah that's my story! :')

Answers inline

Likewise... although I would term mine
clarifications and questions...

Louis Friend wrote:

rfischer@sonic.net (Ray Fischer) wrote in message

Louis Friend <louis_friend@email.com> wrote:

[snip...]

If I did that I would have had to get my hands on
some pretty expensive equipment...

Nope. All you have to do is look at the existing
data and crunch your own numbers using any PC.

Indeed I could do that. However my PC cannot tell me
which numbers to choose, nor can it provide me with
insight with how to interpret and adjust them.
Additionally PC's find it difficult to account for
the political climate those numbers were collected in.

Or you could look at the research reports and look
at the statements made by the researchers and see if
they overlooked anything.

Your faith in the system in admirable and exactly
akin to mine over a year ago... I am no longer
so idealistic in this regard. Pick up a copy
of Steven Pinker's 'The Blank Slate' and you
will get a dose of reality.

Discover. October, 2002.
http://www.discover.com/issues/oct-02/cover/

I have looked at research reports, many of them,
and I found a disturbing pattern... at least
regarding the ABC issue, researchers don't go
out of their way to say 'we were wrong'.

You must not read the same peer reviewed
journal as I do or attend the same conferences,
I see it on a regular basis.

And you would consider that going out of their way?
As a civilian I would of course have a different standard
than you, since I do not encounter those things in my
daily experience... nor do pro-choice advocates who
have been using invalid points years after the fact...
well of course I have no intention of laying the majority
of the blame on scientists for that; the media will do fine.

Be that as it may, I would appreciate a clarification...
you have found the same level of candid retractions
specifically regarding abortion-breast cancer studies,
as for other subjects? I concede my journal research
as yet has been limited to solar cells and the ABC
issue, and that one thing on prostate cancer.

Which I think serves as an appropriate example of
what I'm talking about.
http://www.sciencemag.org/cgi/content/full/282/5387/199a
The first analysis used statistical adjustment and found
RR's of 0.62 (0.25 to 1.55), 0.70 (0.31 to 1.58), and
0.81 (0.36 to 1.80), whereas a re-analysis using case
controls resulted in "RR values were slightly above 1.0,
but all CIs included 1.0, and the lack of any association
between IGF-I and prostate cancer remained."

As far as I'm concerned they messed up with their first
analysis, I'm confident about that because of studies
conducted on this subject... and the fact they didn't
bother to include their full results of the second analysis
in the abstract. I know that isn't a 'smoking gun', but
it is a red flag, or at least interesting. (I'm just trying to
get you to understand why I'm in the mindset I currently
have about scientific, not honesty per say... but being
candid and straightforward about mistakes.)

Perhaps I'm in the dark on this one, but is it standard
practice in science publications to obliquely retract an
errorneous result ~7 years after publication? (even though
it was pointed out in 1994)

(Study): Lindefors-Harris BM et al (1991)
Am J Epidemiol 134:1003-1008
(Admission?/Retraction?): Meirik et al (1998)
J Epidemiol Community Health 52:209-211
Daling et al (1994) JNCI 86:1584-92

Once you bother to get the research reports
yourself and have them in front of you, as I do,
*including* correspondence afterward, I can
hold your hand on it. We may not always agree
on things, but I can guarantee you 3 things...
I can show you blatant mistakes in methodology,
minimizing holes in data and that I've done my
homework.

Get:
IGF-I and Prostate Cancer.
Aug 21, 1998. Science. Mar 5, 2003.
http://www.sciencemag.org/cgi/content/full/282/5387/199a

Melbye et al (1997) N Engl J Med 336:81-5
Melbye et al (1997) N Engl J Med 1835

Lindefors-Harris BM et al (1991)
Am J Epidemiol 134:1003-1008
Meirik et al (1998)
J Epidemiol Community Health 52:209-211

And here is a great article on the power grid for fun:
http://www.tipmagazine.com/tip/INPHFA/vol-9/iss-5/p8.html

Science does not require blind acceptance.
The data, the methodology, the assumptions,
and the conclusions are all made available.

Therefore you shouldn't have trouble acquiring
the above. I am patient and don't expect you
to be ready for a bit.

then YOU get to prove it.

I thought I had done a decent attempt at that
in the previous thread we participated in.

Whining that

Have you considered becoming a editor?
Your objectivity is inspiring. I suggest you
avoid *interpreting* me... because you are
very poor at it, and focus on other matters.

1) you don't like the conclusion, and 2) science isn't
perfect, does not even begin to address the actual study.

True! But I do not have access to the study as yet.
However since it is a meta-analysis, it is based on other
studies... I have access to those and I've found they have
not established a number of things, and their arguments
make clear they don't know the answers, and that these
unknowns have been carried through in this meta-analysis.

Not exactly. In the meta-analyses I've conducted I've found no
relationship between effect size and publication bias. None, nada, zip.
For instance, see http://www.lyonsmorris.com/absstudy/Table3.htm.

I've seen similar results with methods to encourage recycling,
therapeutic effectiveness, and the relationship between personality
variables and hypnotic susceptibility, to name a few.

While a meta-analysis is only as strong as its literature base, any
competent researcher in the area makes sure that the analysis is large
enough to address those concerns

1) How closely rat breast cancer models humans

Not relevant. Look at the animal models of breast cancer, you will find
that the animal models are adequate surrogates. Otherwise the findings
will not be accepted.

I'm not asking for them to be accepted,
(I assume you mean as scientifically valid as human results)
rather for them not to be rejected. They are what
they are, no more, but certainly not less. I feel rat results
are akin to a lie detector for a trial, not admissible but
they should carry some weight (for scientists AND the
public alike)... but it seems to me rat results are being
treated like hypnosis testimony... which can be even less
reliable, or very helpful if properly conducted...

Hmmmm... another failed science analogy? :')
Point being scientists appear to be emphasizing the
unreliability of rat results, rather than the fact they
are indeed 'adequate surrogates' which have a proven
history of leading science in the right direction. I can see
why it's done given anti-abortion emphasis on rat results,
but doesn't it in the end leave an inaccurate view
of rat studies?

2) If the theory of undifferentiated cells in the breast
being more likely to become cancerous is false
3) If recall bias actually exists in the studies they
consider 'bad', and if it could be statistically
significant

Publication bias can be adequately tested. In the five meta-analyses
I've conducted I've not seen any bias. In the other eight where I've
been a paid consultant, there has not any significant variation due to
publication type (unpublished, confernce only, journal publication, or
book publication).

I'm not sure who's at fault here, but your confusing the hell
out of me. :') What is 'publication bias'? And then are you
agreeing or disagreeing with me in regards to 'recall bias',
since you say there is none... but your talking about
publication bias. I haven't even come across that term till
now, let alone considered it.

4) How much hormones influence breast cancer
5) The impact of various confounding factors
and their interaction with hormones
6) Precisely why breast cancer rates are on the
increase, particularly among wealthy women
like those in Marin county, they list factors like
late child rearing (which is certainly true,
sometimes in conjunction with abortion), but they
don't know which risk is from what.

Point is I have been addressing that meta-analysis by
commenting in general terms at the flaws and now
specifically the gaps in knowledge the meta-analysis
cannot and does not address.

Any adequate meta-analysis will address any reasonable objection.

Absolutely, assuming it is indeed adequate to your standards,
which I have little doubt are high and thorough. But
here I'm confident regarding how meta-analyzes can be framed
to 'address' 'reasonable' objections. Whether the address was
sufficient, and the objection reasonable can be contentious and
left unresolved, especially in the domain of the layman.

Remember, all that any statistical analysis does is provide an estimate
of the actual relationship within the population. In other words if you
duplicate the study an infinine number of times (assuming here you use
the alpha =.05 criterion) 95% of your results will fall within a certain
range.

Of which I have the utmost respect, and fairly good grasp of;
however the same cannot be said of the majority out there.
The media isn't interested in margins of error, but in conclusions.

And as I hope I've shown with the prostate study:
http://www.sciencemag.org/cgi/content/full/282/5387/199a
on issues that where a potential link is small...
the confidence interval regularly can encompass the
entire statistical range being debated.

What meta-analylsis does is assume that different research results
provide slightly varying glimpses of the population relationship. By
accumulating these different results, weighting them by the sample size
(so that you give greater weight to studies with larger sample studies)
you have a better estimate of the relationship within the population.

instead of looking at the science.

LOL! I know you are, but what am I.

The validity of methodoly can be determined from the studies.

True.

Those methodologies which do not provide consistent results
are inherently flawed.

Not always true. Since differing data will provide
inconsistent results with the same sound (or flawed)
methodology.

That's a testable hypothesis. I've found in most of my meta-analyses
that methodological adequacies have a significant impact on the effect
size. In other words if there is actually an effect, it is negatively
impacted by poor studies.

Couldn't agree more, if your saying what I think your saying,
which your probably not. :') What is the effect size... the RR
in the studies, or the meta-analysis. And how would one
know it is being negatively impacted at all if the effect isn't
precisely known... or is the knowledge in the fact the studies
are 'poor'?

Larry C. Lyons

========================================================
Life is Complex. It has both real and imaginary parts.
========================================================
Chaos, Panic and Disorder. My work here is done.

Nifty sig.