Statsitically of some value, just hard to interpret
Chris sutor's response shows he doesn't understand the statistical
basis of a case-control study. This paper shows that silver cars account
for a smaller proportion of cars involved in accidents (5.3%) than would
be expected, given they constitute 11.3% of cars on the road. One may
argue about the way the control population was collected, but (as Y Adi
alludes) the colour spread of cars on the road is the best comparator, as
it reflects the population at risk (ie cars being driven) better than just
the colour spread of cars in existence/ sold or whatever.
There are obvious confounders -hitter or hittee in an accident,
metallic paint cross-correlating with safety extras, driver personality or
even driver age/ experience (my first car was black, my second dark grey,
my third black, my fourth green, and my last two have been silver).
However, in contrast to T.D. Rajan's views, there is a clear end
result of such research. Fleet car managers could compel employees to
drive silver cars; insurance companies could offer discounts (surely there
is an actuarial study of car colour about somewhere - I can't believe
insurance companies are blind to consequences of car colour!)
Interesting stuff. Worthy of the Christmas BMJ.
I drive a silver car... but my wife's is dark blue. Should we change it?
Competing interests: No competing interests