Det Gard cy

pl Ot
rent eur.

[x pushol
Lda November 11, 1975

Dr. I.J. Good
Department of Statistics and Statistical Laboratory

Virginia Polytechnic Institute and State University
Blacksburg, Virginia 24061

Dear Dr. Good,

Apropos the postscript to Statistical Fallacies: I wonder if you
chose the most fortunate example about the irrelevance of a retrospective
study. In fact, there is a socially important inference that can be made
from the (factual) finding that the grandparents of idiots are mostly
normal. That is: the probability of normal people having idiotic grand-
children must be close to the incidence of idiocy in the total population,
That assertion would be falsified by a contrary finding, the extreme of

which is self-evident ~ for example if the grandparents of idiots were
always abnormal.

The inference that I assert here is perhaps not thought of as being
as important as a causal claim, and it is certainly a weaker assertion.

However, it may often be the most important finding from a retrospective
study.

I am certainly not attempting to weaken your criticism of the retro-
spective approach, but so much important and useful information has come,
and sometimes may only come, from this technique that I hoped to offer a
somewhat different perspective. The recent werk by Herbst et al. on tracing
the mothers of girls with vaginal cancer and finding that very many of them

had been treated with DES during pregnancy is perhaps one of the most
Striking recent examples,

I am looking forward to the appearance of your article with great
interest and also to any further comments that may develop.

Sincerely yours,

Joshua Lederberg
Professor of Genetics

JL/rr
ec: Dr. William Kruskal

6-4 ‘CoeD