> I am going to have to protest your misrepresentation of statistics
> in your most recent posting on this thread. Your mistake is a common one,
> and must be stamped out repeatedly and vigilantly by statististicians.
>
> Your concept of "causal distance" is well and good for philosophical
> discussion, provided you can define it with sufficient clarity. However,
> statistics, no matter how exact, can only prove correlation, never
> causation.

        Never said it could.

        Your claim of error can only be upheld by assuming that every
        reference to 'relationship' in my original post implied a
        reference to 'causal relationship'.

        I claimed that statistical analysis could establish a relationship
        between two events.  So it can, such a relationship being called
        a correlation.

> As to how correlation is linked to causation, the best answer is
> weakly. Notice that correlation is symmetric, C(X, Y) = C(Y, X),
> and yet one of the fundamental components of the concept of causality
> is a lack of symmetry. (Few would argue that cancer causes smoking).

        Time flies like an arrow.

> It is an all too common occurence in todays quantitative society
> that statistics are called upon as evidence of causation. The
> mantra that should be remembered is: 'Correlation not causation'.

        Are you saying that statistics CAN'T be used 'as evidence of
        causation', or only that this happens too OFTEN?

        I have no argument with statistics being unable to PROVE causation,
        but I would maintain that statistics can provide EVIDENCE towards a
        proof of causation (which underlies my contention that doubt about
        causation can be satisfied by a convincing statistical argument).

___Bob