Tuesday, December 07, 2010

David Lammy shows why lawyers should never do statistical analysis

A few years ago, I watched an episode of Question Time when David Lammy and George Galloway were on the panel. Lammy spent most of the evening taunting George - ironically behaving like the stereotypical Oxbridge Union debater who spends most of the evening badly heckling others, and not being anywhere near as funny or clever as he thinks he is. In other words, he came across as a total and utter prat.

I was reminded of this when I saw Lammy's figures being quoted in the Guardian today. As an example of how not to do it, this has to go down well. Some details of what's wrong with this analysis is here - including making the point one didn't need lots of FoI requests as much or all of the data is made available by the university itself.

Even the Guardian article makes it obvious what is wrong with the analysis:

A spokeswoman for Oxford said: "Black students apply disproportionately for the most oversubscribed subjects, contributing to a lower than average success rate for the group as a whole:
That alone should send up red flags that what is needed here is a proper statistical analysis of all the factors that can affect the success rate of an application. Any competent statistician would start of by collecting data on all the potential factors (including popularity of the course, race, sex, school type, A level results and then run a detailed analysis of this to discover which elements are statistically significant). One case where this was done was in an analysis of the application of the death penalty in the state of Georgia - where it was clearly shown that statistically, the race of the murder was not significant, but the race of the victim was (ie: if the victim was white, the murderer was more likely to be executed, but this was true regardless of whether or not the killer was white). Ironically, in my Open University module where they presented this data analysis, they took it from a book called "Statistics for Lawyers". Perhaps someone could get Lammy a copy!

In other words, all this analysis proves is that Lammy is, in fact, a prat. Anyone who had seen that Question Time show already knew that, but it's always nice to have corroborating evidence.

(Note: this doesn't prove Oxbridge admissions aren't biased. And before the usual suspects make the comments on Facebook, nor should it be taken as a statement by me about the rightness of the death penalty, which I oppose).