[ RadSafe ] Re: "-authors do not report-" data refuting their conclusions!

[John Jacobus]

Again, the left column represents the data in that
range of doses.  The rows represent the extent of the
analysis.  It the bottom row of first section,
considers the analysis of only two columns of data, 0
and 0-9 rad.  The trend test and value P shows how
well the curve fits the data.  As you go up the row,
more data is used to develop the curve, and as you can
see the test of the curve fit gets better through the
whole range of data, even if it gets worst at the
lower ranges.  

I do not know what you mean by the "one-tail test."  I
believe what you are referring to is the chi-squared
test.  However, I believe that they are using some
other correlation test to evaluate the fitting of the
dose to the cancer incident in table 2, which is not
the same thing.  The analysis of the fit results the
P-value, which they discuss under "Results" on page
19.  Since you used to study epidemiology, I assume
you understand their analysis.

The point that I am making is that the observed
cancers does not change in any one column.  However,
the expected value changes as a new curve is developed
for the data used.  In the bottom row of the first
section, the expected and observed cancers are equal. 
One could say that at the 0-9 rad dose levels, the LNT
is a good predictor of radiation effects.  As you add
higher doses, the LNT is less so, which is not
surprising as there is a lot of "noise" data based on
the P value from the bottom row of the first section. 
However, as I have mentioned, the expected is only a
mathematical projection.  While the actual number of
cases may, as you say, not fit the LNT hypothesis,
that is not be the complete story.  For the lowest row
in the first section, it does a very good job.  

If you want to look at the epidemiological data that
compares expected and actual cancer in the population,
you should look at the one page from McGregor and Land
report of 1977 that I have attached. 

My point is that you have to be clear in what you say.
 The fact is that x numbers of cancers were observed,
which is a fact.  However, when you say expected in
this study, the expected is based on intrapolation
from the data.  It is not epidemiological data which
is what you should be comparing the observed with the
expected.  From the McGregor and Land paper of 1977,
the attachment, it appears that there is no beneficial
effect from doses of 0-9 rad.  

When you say that the number of observed is different
from the number expected on the top line, I would ask
what is the basis of you expected number?  This is
good science and not data mining.

-- John

--- howard long <hflong at pacbell.net> wrote:
> John,
> The left column, labeled "Dose range kerma", unlike
> the rows, covers dosage from a given level to 0.
> That is consistent with their one-tail test, which
> assumes that any radiation under the specified
> amount gives increased risk of cancer. 
>  
> Their own top rows of O and E do not fit that LNT
> hypothesis.
>  
> Howard  
> 
> John Jacobus <crispy_bird at yahoo.com> wrote:
> The question is this.
> As you look down the columns of table 2, the
> observed
> cases stay the same. However, the expected number of
> cases change. I say that this is due to the use of
> the data to draw linear curves thought the data
> sets.
> 
> Do you agree or do you have another explanation?



		
__________________________________ 
Do you Yahoo!? 
Make Yahoo! your home page 
http://www.yahoo.com/r/hs
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Pages from McGregor, Land -- Breast Cancer 1977
	-- single page.pdf
Type: application/pdf
Size: 104266 bytes
Desc: Pages from McGregor,
	Land -- Breast Cancer 1977 -- single page.pdf
Url : http://radlab.nl/pipermail/radsafe/attachments/20050311/40ceccfc/PagesfromMcGregorLand--BreastCancer1977--singlepage-0001.pdf