statistics and lies.

we could set out an entire website on the misuse and abuse of statistics. some of them involve drawing spurious correlations from sparse data sets.

but then there is also stupid probability and lies dressed up as something interesting, or as page fillers. what not to do if you are a journalist:

Excerpt from the Daily Mail, about the FA Cup draw

"Once Clemence had plucked Blackburn from the bowl, the odds were stacked hugely in the FA's favour. A 66 per-cent chance of getting the winner from the two remaining replays. Only a 33 per-cent chance of drawing poor old Watford next."

This is absolute rubbish. Blackburn had a 66 percent chance of facing Man Utd or Chelsea before the first ball was drawn. The fact they were the first ball to be drawn changes nothing.

To illustrate this, say Man United had been drawn first. What is the conditional probability that Blackburn would face them. 1 of 3 remaining balls was blackburn, thefore 33%.

If Chelsea had been picked, the conditional probability is 33%. If Watford was picked, also 33%. Sum up the conditional probabilities and they add to 1, the daily mail has taught us nothing new with that statement.

Still, there are many desperate and smart Russians around who would tell you the same thing. It is interesting how many postgraduate scholarships there are for someone of Eastern European origin

No comments: