Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: causation



I have just recently achieved a small insight into
the difference between a statement about _statistics_
and a statement about _causation_.

*) Statistics allows you to verify an implication,
A implies B
and this naturally leads to the prediction that if you
draw more samples from the same distribution, you will
continue to observe B or not A in future data.

*) A statement about causation makes a much bolder
prediction. The hypothesis
X causes Y
means that even if you intervene so as to change the
distribution, you will observe either Y or not X in
samples drawn from the new distribution.


So one of the key ideas about causation is that even
if you change something, you cannot get rid of the
effect without getting rid of the cause.