Re: Cause and effect

[Tuan V Nguyen <t.nguyen@garvan.unsw.edu.au>]

Hello anh AiViet, Huy, Vu and folks, 

>  I think Cause and Effect is the Raison D'E^tre of Statistics even it is 
>not a problem of Statistics as Anh Huy said. The most beautiful thing of 
>Statistics is that it changes the way we think of Causality.

	Let me take a concrete example: If your wife (or lover) 
feels happy after you gave her a red rose, then you may say 
that the rose is the cause of the effect of her happiness. 
However, if you give her a rose, and if at that precise 
second, a piece of toast pops out of the eletric toaster, 
then it would be ludicrous to make any inference regarding 
the rose and the toaster. 

	The distinction of cause and effect has been a subject 
of discussion among statisticians for quite some time. One of 
the main domains of statistics is the study of relationships 
between attributes or variables. So, when one writes the 
equation Y = F(X) + E, many readers immediately think that X 
causes Y. But, of course, the inference can never be complete 
without a logical reasoning of the phenomenon under study. 
Consider the equations

	WEIGHT (in kg) = -12 + 0.5*HEIGHT (in cm),

and
 
	WEIGHT (in kg) = 68 - 0.04*Age (in yrs)


The standard interpretation of these equations is that (1) if 
you can increase your height by 1 cm, you are expected to 
have your weight increased by 0.5 kg; and that (2) if you are 
celebrating your birthday tomorrow, you are expected to drop 
0.04 kg in weight. But, of couse, this is only a 
relationship. There is no biological evidence suggesting that 
increase height will CAUSE increase in weight, nor is there 
evidence suggesting that age causes decrease in weight. What 
we can say is an ASSOCIATION between height and age vs. 
weight. In fact, a lot of relationships between phenomena can 
be classified as ASSOCIATION rather than CAUSATION.

	A few weeks ago, there was a report that a certain drug 
could reduce the incidence of cholera in Vietnam. The finding 
was based on a study in which half of patients received the 
drug (treatment) and another half did not receive the drug 
(controls). What they actually found was that the incidence 
of cholera in the treatment group was significantly lower 
than in the control group. Based on this, can we say that the 
drug caused reduction in risk of cholera? Having worked with 
medical fellows for some time, I must say that I even doubt 
whether the immunization experts can answer this question 
properly. Statisticians have invented a wonderful word for 
this; they would say something along the line "the drug was 
ASSOCIATED with a reduction in risk of cholera".

	In the last 30 yrs or so, billion of dollars have been 
poured into genetic research, and despite some laudable (or 
laughable) claims from medical researchers, we are still at 
dark regarding mechanisms of major genetic diseases. The 
inter-dependence among organs in our body is so complicated 
that it is thought impossible to make any inference on 
causation. What we can say at most is association.


	Tuan V Nguyen