Gary Klass (Illinois State) has posted his paper, “Just Plain Data Analysis: Common Statistical Fallacies in Analyses of Social Indicator Data” to the political methodology working paper archive. It’s an easy to read paper, in which he breaks down statistical fallacies into measurement and construction of social indicators, and those due to causal inference. Klass concludes by making a couple of good points:

Identifying and avoiding statistical fallacies requires subject matter expertise more than trainable statistical skills. It requires a good understanding of probable alternative explanations for numerical findings and a good understanding of the social indicators employed. Familiarity with the indicators requires an understanding of how the data have been collected, how the indicator is constructed, and how readily available alternative measures might produce different results.

These points are worth remembering as we head into what looks like another very close presidential election; I suspect we will hear a lot of argument and debate about all sorts of statistical analyses and inferences this fall.