Posts Tagged ‘Crimes against statistics’

Pundit has a post up by Rob Salmond looking at the trends from a linear regression to recent poll results, which they then extrapolate to the election.  They have National winning 50.8% of the vote at the election, Labour, 24.8% and the Greens 14.6%.  The results are pretty much within the margin of error of the predictions here, but the whole concept of doing a linear regression and extrapolating the results forward is a bit sloppy.  For some reason it just reminds me of the statistical analysis I’ve come to expect from The Standard.

There is a lot wrong with this approach, but instead of going through it I will just quote another blog post from earlier this year.  From Pundit.  By Rob Salmond.

When a pundit says “if this trend continues to the election,” stop listening

The original piece was a must-read.


Read Full Post »