Medicine's Ten Greatest Discoveries

David Lloyd-Jones dlj at
Sun Sep 6 04:38:14 EST 1998

Steven B. Harris wrote in message
+AD4-In +ACI-David Lloyd-Jones+ACI- writes:
+AD4APg-In somewhat the same sense, multiple regression analysis, which was
+AD4APg-destroyed epistemologically by Locke and Hume a couple of centuries
+AD4APg-it actually came into general use, has no claim whatsoever to Platonic
+AD4APg-logical rigor -- but is one of the most useful forms of explanation
+AD4-that we
+AD4APg-have.  Usually right, on rare occasions wrong.
+AD4-   In your dreams.  Logistic regression is mainly good for ruling out
+AD4-causation (in which case it really is usually right, and only on rare
+AD4-occations wrong).  But for cases where variables correlate, the usual
+AD4-case is that direct causal relationships cannot be inferred, and quite
+AD4-often do not exist.  Rather, correlations are usually caused by third
+AD4-factors which are causal to both variables, which have no actual
+AD4-relationship with each other.  Such confounders or proxy variables
+AD4-bedevil every multiple logistic regression ever done.  If not,
+AD4-prospective studies and experimentation itself would hardly be
+AD4-necessary.  All we'd have to do is epidemiology.  If black people score
+AD4-less well on intelligence tests, why then, being black must cause brain
+AD4-problems.  If gay men use a lot of nitrite recreational drugs and get
+AD4-AIDS, and the more nitrities they use the more likely they are to get
+AD4-AIDS, why then nitrites must cause AIDS.  If married people live
+AD4-longer, why then marriage must make you live longer.  If people who jog
+AD4-every morning live longer, why then jogging every morning must make you
+AD4-live longer.  If there is more juvanile crime during those months in
+AD4-which ice cream sales are highest, it must be that ice cream causes
+AD4-deliquency.  And so on and so on, into foolishness ad infinitum.


The fact that there is conspicuous foolishness in some corners of the world
does not prove that it is not exceeded by the wisdom and intelligence

The examples you list above are all famous ones, of the most crooked,
stupid, pre-judged, and bigoted misuses of shallow numerical methods
anywhere -- where they were not, like your ice-cream one, simply made up by
stats profs to demonstrate a point.

You are quite right that negatives of correlations are  probative, while
positives are not. Positives, however, are very often right, often
surprisingly and usefully so -- as with your good example of epidemiology, a
science more useful than say surgery or curative clinical medicine.

You don't need a doc if the bailiff took the handle.


More information about the Bioforum mailing list