Last night I saw Jon Stewart interview statisticians Martin Gilens and Benjamin Page who have done an analysis of the influence of interest groups on US politics. The first part (seen above) was shown on Comedy Central and is not Stewart's best interview. The other web extended interview parts are shown below and Stewart does a better job. The challenge for statisticians is to explain their results in plain English. That might be the real secret to Nate Silver's success.
Multivariate analysis, for those of you who are unfamiliar with the term, is an analysis where there are multiple outcomes of interest and one or more predictor variables. A recent example of this is when I did a factor analysis of state health, income and the concentration of hate groups to which variables were most highly correlated with hate groups. Hate groups were strongly associated with the health variables rather than the income st the state level.
Multicollinearity is a related concept. In a multiple linear regression, you have several predictor variables and one outcome variable. Multicollinearity occurs when you have clusters of highly correlated predictor variables in your model. This can severely skew the terms in your model.
There is a mountain of data which page and Gilens sifted through to assess the influence of money on politics. There is already a mountain of research on this topic. They were taking a more comprehensive approach and found more specific results.