Bliley PM
VII. Overprecise and Unknowable Figures; [and] VIII. Semi-Attached Figures
Abstract
Contains draft chapters from Huff: "VII. Overprecise and Unknowable Figures" and "VIII. Semi-Attached Figures". Defines "overprecise and unknowable figures" as "any conclusion based on faulty or incomplete information or upon an inference of causality where none has been demonstrated". References summary statistics on smoking and health as exemplifying concept. Defines "semi-attached figure" as a concept in which conclusions do not directly follow from the available data. Critiques animal smoke inhalation studies in terms of their applicability to human smoking situations. Quotes from published studies and popular culture, and uses didactic methods. Relates to Bates 2025037587 (see "Document Quotes" field for excerpts and Bates 1005087402).
Fields
- Quotes
Remarks "[i]t is not difficult to combine an estimate or assumption with an arbitrarily chosen method of calculation to produce any conclusion wanted".
"Inconsistent reporting at the source often produces medical statistics of the kind that might be called semi-attached." (from Bates 2025037423).
Observes "[m]uch of the alarm about lung cancer has been produced less by recorded cases or deaths than by extrapolations to a decade or two in the future."
- Type
- Article
- Draft material
- Position paper
- Author
- Huff, D.
- Named Person
- Auerbach, O. Dr.
- Brothers, J. Dr.
- Engel, L.
- Ferguson, G.R. Dr.
- Finkelstein, H. Dr.
- Flick, J.B.
- Fuchs, V.R. Dr.
- Garland, L.E. Dr.
- Garvin, C.A. Dr.
- Kennedy, R.F.
- Kinsey
- Lincoln, A. Pres.
- Marx, K.
- Moncrieff, R.W.
- Pauling, L.
- Porter, S.
- Reichmann, W.J.
- Seligman, Robert B. (PM VP of R&D c. 1976-82)Vice President of Research and Development at Philip Morris Richmond, VA 1976-1982. Reported to Senior Vice President of Operations. In 1982 transferred to tobacco technology group. Wanted to share ammonia and other tobacco technology with PM International companies.
- Taylor, H.C.
- Twain, M.
- Wagner, R.F.
- Ward, A.
- Named Organization
- American Cancer Society
- American Petroleum Industries Committee
- Atlantic Monthly Magazine
- Attorney General
- British Ministry of Works
- Bureau of the Census
- California State Senate
- Citizens Committee to Keep New York Clean
- Colliers Magazine
- Delaware Medical Journal
- Department of Agriculture Agricultural Marketing Service
- Engineering and Science Magazine
- Federal Bureau of Narcotics
- Fortune Magazine
- Geigy Agricultural Chemicals
- Harpers Magazine
- Journal of the Medical Society of New Jersey
- Mattachine Society
- National Board of Fire Underwriters
- National Bureau of Economic Research
- National Safety Council
- New England Journal of Medicine
- Russell Sage Foundation
- San Francisco Examiner
- Sanford University School of Medicine
- Surgeon General
- Surgeon Generals Advisory Committee
- This Week Magazine
- Time Magazine
- True Story Magazine
- University of California
- US Dept of Agriculture
- Yale University
- Region
- United Kingdom
- United States
- Keyword
- Emphysema
- Inhalation studies
- Kinsey Report
- Life on the Mississippi
- Lung cancer
- Smoking and Health
- Surgeon Generals Report
- Use and Abuse of Statistics
- Subject
- Cigarettes
- data analysis
- Diseases
- epidemiology
- Federal level
- Government agencies
- Health advocacy groups
- Health effects
- Human subjects
- industry sponsored research
- Regulations
- Research studies
- Surveys
- Animal subjects
Document Images
~ 0VERPRECISE AND UNKNOWABLE FIGURES
!-..._"It ain't so much the things we don't know that
et us in ~rouble. It's the things we know that.aln, t
" said the American humorist, Artemus Ward. .~-~.... ".:'~
SO,
~.~iii He might, of course, have b.een thinking of~. any.
onclusion based on faulty or incomplete information. .
or upon an inference of causality w'here none has been
demonstrated. But his words apply most aptly to a .
special kind of statistical mistnfo~atlon--the ira-.
possibly precise fi~e. .:~,: .:.. ~-. -
The insidious t~ng about overprecision is that
.the precision itself is what makes such a fi~e so
.... impressive, it somehow robs us of the ce~on sense
that should tell us that the fig~e in question is
one that no one could possibl_
hess or even, in many cases, know. even a~rox_mate-_.
OCR for 1005087596 does not yet exist
"statistically significant at the 5 percent level "
This is found on Page 86 of Smoking and Health and
i, w~s-obtained by combining the results of four studies.)
~ ~.. It should be n'oted in passing that combining
from several studies not conducted in the same
is questionable procedure, and the advisory com-
~~_. ; .,..' .~. ~.::~ '~.~ ~,'., , ~,~ 7,--~,~i~.'_~.~'~;~'.
~.,,. ~.....~,,~ ... ~,_ ~,~~, ~ ~,~. .... - . • ~ ,. ~ ~-,~_ ,~ : ~ : ....... .-,: .;~
-,~-
to the S~geon General recognizes this. -..
': :. ~"]:~However, what we are concerned with here is ex-
. pressing the Comparison between the groups in a
':mortality ratio (1.20), and this is a perfectly proper
--~ _~ ~ ,~.,-.~.~ , .~
~-~:~.~':thingto do. But it does have an ~fortunate result: .........
"
~.~.~, • ~.-, ~ , . ~ .
"'~it m~es it appe~ that we now ~ow the actual mor~ality
ratio of two kinds of groups right down to a decimal ....
place. The reader must bring to his interpretation of
~ .~ this fig~e a ~nowledge that what looks like a rather
~-~-..,~-~ exact fi~e is only ~n approximation. From the
,~:.,<-:., accompanying statement of significance ("5 percent "
~~ '"" level") we discover that all that is actually ~o~
.'~%~i'~:'~;. " is that the odds ~e 19 to one that the second group
~ ~..~..
" :.~~ truly does have a higher death rate than the first.
.,.'L~;~-~~,~,~.~.,<~ .. • ..,..~ The actual increase from one group to the other may
.,,.~::-... be much less than the 20 percent indicated, or it may
.-. ~ be more. And, of course, as has been more fully de-
, .-~-.~ ... veloped in the earlier section on s~ple size and
~:~ signific~ce, .there is always the 5 percent chance
~~:~,"' " ~ " that n0 difference at all exists in the direction - ~
#- overprecision
_. ": ; ."L • .- " - . :..,
Indicated. That is, a larger sample might (one chance
~i:.:i 20) have 'shown the death rates to be about"the Same
,in bothgroups or greater .in the first group than in the
• ~ ~ ,.~'--" ~ t ~.',.~ ~, , ~ . .
econd. .,i -.~~-~..'~."..~.~,~.;::'; ~ii,: ~.~. " • "
papers and a news maga=ine published a report
.some yearsago that the average Yale graduate of the
lass of 192g was earni $2 ear, which was
~-~ ,~,~,~', . ~ ,~-.. ~.~. . j'.,~';.~." ,.~... _ . ...
even more money then than it is now. No doubt there
some pretty poor statistical .work behind this
~risingly generous figure. Since it was derived
~ guess that a
the sample by not replying. It is a -~
disproportion of those with low incomes would be in
both these groups.
But note how impressive it is to read that
the average income is $25,111. This lends an aura of
accuracy that would be lacking in a statement tha~ said
"about $25,000." Yet obviously that figure was obtained"
by averaging a lot of round figures few of which were
a~vthin~ like so ~rec~se as that themselves.
.necessarily omitted from the ss~mple; and those who
were reluctant to reveal their ~ncome status to their
old classmates would tend to remove themselves from
replies to a mail questionnaire, it is reasonable
.. :~.~ - ...~ !.~. . . .
,~- ~~:.~i..~,;~'
to suppose that it was primarily a produc~ of sampling
, ,,
. ,.. ,
.~bias...of selection and self-selection. Men whose ,
~N:~-~t::;,~-~: -:-~addresses were ~nown to the al~]mni office were .
~(study in New York that once reported that a working woman
-,,-needed a weekly pay check of $#0.I} to get by. This
-conclusion does not convey any information tha$ could
be fo~d in a statement that she needed "~o~d $$0 ~.L'~
week." Yet somehow it seems to. "
.~.The ~erican Petrole~ Ind~tries Co~ittee issued
statement some years ago that the average 'yearly t~
for automobiles was $51.1~. -
~. " The NatiOnal Board of Fire Underwriters reported
~,~,~ ...... ,.'total fire lesses for ene year--1960--as $1,107,82~,000.
~:i~:.:~;:::~~ ?~,:;,;; By blithely offering a figure down to the last thous~d
~-~;~,"~.~-~~,~-~::~doll~s, the Board is implying that it ~ows the extent ~'
.,%?i;:,~r::-..::.~ i/~:. whole t~ng becomes even less impressive when ~he method
-" " .... " " • settlements but the initial claims. And to this, it
-" adds an estimate of ~insured losses So what lies
.... behind the figure it publishesis two things:
" exaggeration of ins~gd losses and an apprcxlmation
of ~insured ones. But you'd never ~ess that from
~ """
~ _
~ ~,~_."~.-
- - the confident total that is published.
. ~'~ ~ .'.~.~
,.~ :.. Some such method as this lles behind m~y of
~ . ,- the overprecise fig~es that ~e produced each year. ~'~ :-~-~,.~:;~..~L:-
6- overprecision
OCR for 1005087601 does not yet exist
8 overprecision
per cigarette smoked."
i.~:i It is not difficult to combine an estimate or
Lassumption with an arbitrarily chosen method of calcula-
,n to producealmost any conclusion wanted.
~'~~:~.. For a year (19~9) in which the Bureau of the Census
' •
~ald that the "income Of the average family was $3,100,
ithe RuSsell Sage Foundation described it as $5,00~.
From the precision of the latter figure~ it would
appear to the casual reader that the Foundation knew
what it was talking about. The government
looks like what it is--an approximation.
..., ,,~,., . ~ ..... it indicates merely the neighborhood in which the truth
• -":~'::.~'.~.::.,.':~.":.,:-:'probably lles. As for the s~prisingly exact fig~e,
:~:&.~:::.::;i;... ..... -~~ it is a product of some odd ari~etic The Fo~dation
~ ".:::" " began with the total personal income of the ~erican
:~;:~:(- '~ people~ divided tha~ by the estimated population ~o get
~.~ .... an average of $1,251 for each person. Multiply this
by fo~ the very approximate size of the average f~ily~
When these figures are pursued to their sources, ..
~wever, something quite different about their relative
merits emerges. The round figure is a median, produced
..~ii;/~7;>i:;~:;i'.':7-:.,~:;'!i:: careful sampling study--and then rounded rr to "
"J ....:" ........ ..... n a user that, like all such statistical dezivavions,
,r.!:~.:;.,~, .: ,; ..-;. ,-...j.. :, . war
'-.:..-..L ~.-'.~
and you have that $5,004.
That personal income figure, by the way, is anothe~
of the things that is never known with anything like the
precision with which it is expressed. The same can be
. . •
..~,~,~ ~....
said or another big figure that is watched with eagle
eyes by.buslnessmen, economists, and gover~ent people.
This is the grass national product, well ~own as GI~.
• :Of it, Dr. Victer R. ~chs~
Ecenom£c Rese~ch~ has said flatly that It is becoming
'increasingly less useful" as a meas~e of a~thlng
~ instance of why the business writer Sylvia
Porter calls G~ a "spectacular ~dersta~ement~" consider
a t~ical ~erican c~ught up in the de-tt-yo~self move-
:,:.'ment of the last decade. He has bought an old house,
which he plans to remodel extensively over a period of
....... years. Te p~ the cost of hiring carpenters ~d pl~bers,
he increases his earning in the first year of home owner-
...... ..-. this year will not be measured, ~d will not add ~o any
",~. reckoning of gross national product. For reasons like
-L. this, G~ fig~es for one year are not closely com-
~ .-
..~.' ." p~able to those of ~other. And when subst~tial social'-
-..~ and economic changes are occ~ring in the nation, as
the Berkeley campus is about 27,500; and this number
was divided by ten, seemingly without its having
occurred to anyone that mor~ than one third of the
students were women.
The circularity of the logic used to substantiate
the subcommittee s accusation should be noted. The ~:-" . .,"~:~.~'~~-'.~
