Re: Dose-response threshold from A-bomb survivor data

[Bernard L Cohen <blc+@pitt.edu>]

On Wed, 17 Dec 1997, John Goldsmith wrote:

> Professor Cohen:
>      You may have come to a reasonable conclusion, but the use of Gaussian
> assumptions on difference data can lead to instable results.  The Y values
> you use are differences between observed and expected cancer ( presumably
> rates ).  This is confirmed by the negative Y for the lowest exposure.
> This problem can be obviated I think, by using actual observations. I
> suggest you consult a good mathematical statistician at Pitt.  
>                     John Goldsmith  gjohn@bgumail.bgu.ac.il


	The difference shows up only as the centroid of the gaussian
distribution, and has nothing to do with the distribution itself whose
width is determined by the number of actual observations (observed
deaths). I have handled mountains of experimental data for the past 50
years and therefore feel completely confident in this application without
consulting mathematical statisticians.