>an experiment done by sociologists in the 50s, in
>which they got a bunch of stats together which demonstrated that there
>had been a statistically significantly greater amount of people
>cancelling their tickets on flights that subsequently crashed than on
>normal, successful flights. That sounds like an interesting approach.
Trains, as I recall, not planes. The results were provocative but, as
usual, ambiguous and iffy. (As I recall, the comparison stats were for the
same trip in the weeks immediately before and after the event, or maybe for
the equivalent trip the year previously. Weather variations and all manner
of other stochastic wiggles blur such cfs - and presumably the number of
passengers will drop off a bit in the week *after* a fatal crash. I even
found such an effect in Lotto guesses - after a given number has won, it
becomes slightly more unpopular for the following couple of weeks. It's
had its turn, after all. Lightning never strikes... If the odds were
fairer, this would be a useful item of information for rational people
wishing to place bets with higher-than-strictly-correct returns.)
These days with computerised booking it would be very easy for airlines to
do such comparisons. They might even be doing so, but the results would
never be published - PR nightmare! (Admitting that planes crash, I mean.
Not on *our* airline, sir!)
Damien Broderick