CORRELATIONAND

CAUSATION

Somecautionsmustbe notedat this point.

First, you cannot determinethe

causeof the relationshipfromthe correlationcoefficient.

Two variablesmay be

highly correlatedfor oneof threereasons:

(1)

X causesY,

(2)

Y causesX, or

(3)

both X and Y are causedby somethird variable.

Awell known storythat

illustratesthe danger of inferring causationfroma correlationcoefficient between

the numberof storksand the numberof births in Europeancities(that is, the

morestorks,the morebirths).

Insteadof issuinga dramatic announcement

supportingthe mythical powersof storks,further investigationwas carried out.

It

was found that storksnestin chimneys,which in turn led to the conclusionthat a

third variable was responsiblefor the relationshipbetweenstorksand births -

sizeof city.

Large citieshad morepeople,and hencemorebirths; and more

houses,and hencemorechimneys,and hencemorestorks.

Thus, attribution

causality is a logical or scientific problem,not a statisticalone.