Wednesday, 8 January 2014

Haidt, Libertarians and Statistics

In this talk Jonathan Haidt talks about research into the views of Libertarians as compared to Conservatives and Liberals.

The results seem to me to be seriously flawed.

The first point is that at no time does Haidt present any visual indication of the sampling errors in the data.  There may be a difference of a few percent between Liberals and Libertarians in one of the charts, but what significance does this few percent have?

The second point is that there was no indication in the data of the distribution of views.  Were Libertarians all just a bit lower than Conservatives in one of the charts, or were the views widely distributed, or in clumps?

The third point is that no control questions were asked - some of the questions were designed to seek out specific values (or lack of values!) of Libertarians, but no results of neutral questions were presented to see if there was any sampling bias.

The final point is that the results of the most data seems to be a general 'so what?' - the differences between Libertarians and either Conservatives and Liberals was so slight as to make the drawing of  any conclusions seem rather a waste of time.

What Haidt has presented here are slight and obvious effects, with no surprises, and with little in the way of solid conclusions.

No comments: