Discusses the innovative log-linear model of statistical analysis. This model makes no distinction between independent and dependent variables, but is used to examine relationships among categoric variables by analyzing expected cell frequencies.
Though I love this series, I don't know that this ranks among the best at explaining log-linear models. The real problem is that the rest of the series really does an amazing job of explaining ridiculously hard concepts, where as this one was ok, but not fabulous.
I would think, though for those that need to learn about log linear models and are specifically doing work on polling, this would speak volumes to them.