Re: Glass-Based Radon Measurements (fwd)

[Kai Kaletsch <info@eic.nu>]

Dear Dr. Lagarde,



Thank you for answering my questions. Obviously, I have to leave it at that,

since I cannot check the calculations.



I have graphed the categorical ERRs for the surface based measurements and

the ERR x 2.25 for the air based measurements (see

 http://members.shaw.ca/eic/ERR.pdf ).

To me, the two curves are as identical as they could possibly be,

considering the inherent granularity of the data and the fact that the ERRs

are only given to 2 significant figures. I found that curious, since the

shapes of the raw odds ratio curves are completely different between air and

surface based measurements. One is saw tooth the other is essentially a step

function

( see http://members.shaw.ca/eic/odds.pdf ).



Best Regards,

Kai



----- Original Message -----

From: "Frederic Lagarde" <Frederic.Lagarde@imm.ki.se>

To: "Kai Kaletsch" <kai@eic.nu>

Cc: <epirad@mchsi.com>; <radsafe@list.vanderbilt.edu>

Sent: Wednesday, October 16, 2002 12:38 AM

Subject: Re: Glass-Based Radon Measurements (fwd)





> Dear Dr Kalesh,

>

> I do not really understand what it is you ultimately want to find support

> for? However, I will answer to your further speculations, but I will

> recommend you read the detailed methodology described in the paper and I

am

> sure this will clarify any further queries you might have.

>

>

>

>

> > I also found that both of the relative

> >risk curves (air based and surface based) were extremely smooth. (see

> >http://members.shaw.ca/eic/odds.pdf ) Because of the limited number of

cases

> >and controls in each exposure category, one would expect about a 25%

> >statistical variation around the true relative risk curve.

> >My explanation for the smoothness of the curves for the categorical

relative

> >risks, was that one relative risk per 100 Bq/m3 was calculated for the

> >entire data set and that this RR per 100 Bq/m3 was applied to the

exposure

> >categories.

>

> As described in the paper, these are category-specific odds ratios. These

> are of course estimated independently of the regression slope reported at

> the right hand of the table, otherwise there would be no point. You

> complain about a lack of statistical variation about the true relative

risk

> curve. Do you know what the latter is?  I do not , I can only report the

> category-specific odds-ratio with their confidence intervals which

indicate

> their statistical variation.

>

> >If my guess is not correct,

> >do you have an explanation for the extreme smoothness of the curves?

>

> Perhaps if you had not unappropriately drawn line segments to join the

> point estimates within categories, the response would have appeared less

> smooth to you. Actually, it is customary when estimation is based on

> categories to plot the odds ratios as a step function, or to simply plot

> the estimates with their confidence intervals against  category-specific

> exposure averages.

>

>

> >Also, the air based and surface based RR curves have exactly the same

shape

> >and differ only in magnitude. If one multiplies the categorical excess

> >relative risks for the air based measurements by 2.25, one reproduces the

> >categorical surface based ERR values in all categories. I think that it

is

> >very unlikely that this is coincidence. There has to be something in the

> >methodology that forces it. What is it?

>

> The ratios between category-specific odds ratios for the two methods are

> 2.14, 2.29 and 2.26, respectively. The methods are independent but both

> measure radon exposure for the same subjects. Air and glass-based

> measurements are of course correlated (see figure in manuscript). We

> mention in the paper that glass-based odds ratio estimates were about

twice

> those based on air measurement. If one of the methods leads to

> understatement of the risk it is not surprising that underestimation would

> occur to similar extent for different exposure categories.

>

> Sincerely,

>

> Frederic Lagarde

>



************************************************************************

You are currently subscribed to the Radsafe mailing list. To unsubscribe,

send an e-mail to Majordomo@list.vanderbilt.edu  Put the text "unsubscribe

radsafe" (no quote marks) in the body of the e-mail, with no subject line.

You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/