Statistical nightmare
On BBC Breakfast News just now they were talking about a new report on mental health provision. The presenter, Suzannah Reid, went through a number of “damning” statistics in an incredulous manner. Included in the list was this gem: “46% of health trusts are below average performance.”
Hmmm….
So, 4% more health trusts are at or above average performance than would be expected then, if the distribution was normal.. sounds like a positive rather than a negative to me. (Though there would have to be a few *very* low scores to skew the average.)
Why can’t journalists understand the very simplest statistics?
Average Performance = 50%
Comments
3 responses to “Average Performance = 50%”
-
re: Average Performance = 50%
Actually, you could easily get 46% below average even if the distribution is normal. In fact you might expect it. Remember, with a normal distribution, the majority of the data points are going to be at or near the average. So, if you had 12% exactly at the average and the rest equally and symetrically distributed one each side, then 46% will be below average. No skewing required.
-
re: Average Performance = 50%
You’re quite right, I was being sloppy – unless you are talking “median” average in which case “50% = average” by definition – but that numbers in the range 50% +/-5% get cited as being significant deviations from an unspecified expected “average”, plus that it is described as “46% below average” as opposed to “54% at/above average”, appalls me.
-
re: Average Performance = 50%
For me, a simple person with no statistical leanings, it means that I have almost an equal chance of getting below average treatment when I get unwell or something happens to me. Statistically, as an individual, I have only one chance of throwing the coin. Bloody comforting…
Leave a Reply