Tab K


Memoranda concerning sampling of
experiments

-    Memorandum from Duncan Thomas regarding proposed scheme for
     sampling experiments


                           DISCLAIMER

The following is a staff memorandum or other working document 
prepared for the members of the Advisory Committee on Human
Radiation Experiments.  It should not be construed as representing
the final conclusions of fact or interpretation of the issues. 
All staff memoranda are subject to revision based on further 
information and analysis.  For conclusions and recommendations of
the Advisory Committee, readers are advised to consult the Final
Report to be published in 1995.


MEMORANDUM

TO:      Members of the Advisory Committee on Human Radiation
	 Experiments

FROM:    Duncan Thomas

DATE:    July 19,1994 (Prepared on June 3,1994)

RE:      Proposed Scheme for Sampling Experiments



1.	This note is aimed primarily at the medical experiments, since
	the number of instances of intentional environmental releases
	appears to be too small to require any form of sampling.

2.	I am operating under the assumption that there may be many
	experiments that potentially fall under our purview, as 
	defined in the companion document, and that we will not wish
	to describe each of them in detail.  Instead, we will aim to
	characterize the principal types of experiments, provide an
	indication of the extent of the problem, and then describe
	a few representative or extreme examples of each type.  
	Should it develop that the number of experiments is not so
	large, then this proposal would become irrelevant.

3.	A truly scientific sampling scheme would begin with a 
	definition of a sampling frame of all possible observations
	and then select a random sample from it, possibly stratify-
	ing on various aspects to ensure adequate representation
	of all relevant features.  This is clearly impossible in 
	this situation, since the universe of all medical 
	experiments is difficult to identify because of the secrecy
	aspect and in any event would require unreasonable effort
	to assemble.  The purpose is therefore to help us 
	understand the main dimensions we should consider for
	organizing our report and to prioritize the search for
	information.  A secondary purpose might be to provide some
	summary data for the report on the frequency of ethical 
	abuses, should it prove helpful.

4.	Because of the difficulty of identifying secret experiments,
	this proposal does not replace the need for staff to
	continue its detective work chasing up leads.  Nor does it
	replace the need for the committee to provide on-going
	guidance to the staff and agencies about priorities based
	on their general background knowledge or intuition.  I
	anticipate that the more formal statistical data collection
	be done in parallel with these other activities, with the
	aim of providing interim guidance about classification of
	experiments and priority targets for discovery within the
	next few meetings.  Should this prove rewarding, the data
	base might continue to be expanded as the work progresses
	and perhaps used in the final report, but otherwise could
	be abandoned with little effort wasted.


				1

5.	I therefore suggest that the staff begin assembling a
	computer data base comprising all medical experiments
	that have so far been identified, providing summary data
	on each for the various dimension defined in the companion
	document.  The following sources of information should be
	included:

	a.  	The various agency records retrieval processes
	b.	The DOE helpline
	c.	Media accounts
	d.	Advocacy groups, open literature, selected
		institution records, etc.

6.	Some guidance regarding the scope of studies to be included
	will doubtless be needed from the subcommittee to avoid an
	undue burden of staff effort entering data on experiments
	that we have little or no interest in.

7.	Once a reasonable amount of data has been accumulated from
	all sources (even if known to be incomplete), some initial
	tabulations along dimensions to be chosen by the 
	subcommittee should be produced.  This would lead to a 
	first cut at a system of categories that might be used for
	the report.

8.	Once we are agreed on a tentative system of categories, I
	propose we randomly select a few examples of each of the
	larger categories and obtain further information on each.
	This might be supplemented by additional nonrandom samples
	that are of particular interest (e.g., those already well
	documented, particularly egregious examples, secret
	experiments requiring further digging).  Study of the 
	resulting documentation might suggest further subdivision
	or rearrangement of our categories, a refinement of our
	priorities, or a selection of further random or nonrandom
	samples in an iterative process.

9.	It is not clear whether any statistical data will be 
	needed for the final report, but it is conceivable that 
	the committee may wish to make summary statements, such as
	the frequency of ethical violations in different periods
	or the level of risk for compensation purposes.  If such
	statements purport to represent the typical standards of
	practice, then they need to be based on some defined
	sampling scheme (even if flawed) rather than on simply
	anecdotal evidence.



				2