Tuesday, March 04, 2008

Lies, damned lies and (shown with) statistics

Bishop Hill has decided to have a lot at MPs' expenses and see whether he could analyse, statistically, whether they were on the fiddle.
My hypothesis was this: if MPs are employing lots of staff, their office costs should be inflated too, to reflect all the work done by the staff. I therefore prepared a scatter plot of office costs (columns 3, 7, 7a and 8 on the PDF file) against staffing costs (column 4). Here it is:

I've asked Excel to calculate a linear trendline, which you can see on the graph. And if you were in any doubt as to how good a correlation there is between staff costs and office expenses, the answer is that there is none. Literally. (For those who aren't statisticians, the R2 value of 1E-5 which is to say, near as dammit zero, is the figure which tells you whether there's a correlation or not. A value of near to 1 is a strong correlation. Zero means there is none.)

Which strongly suggests that quite a lot of our elected representatives are on the fiddle.

Colour me utterly unsurprised.


Anonymous said...

Ever thought of doing a Benfield number analysis of this?

its usually a good revealer of made-up numbers

some random stat guy said...

Nice how it almost maps the the UK geography. Not the first time this has been seen either.

The Cuban Health System

The Cuban Medical system. Over at the ASI blog, Tim Worstall asks if the Cuban Health statistics are true . I can't comment as to th...