From Schrodt’s “Seven Deadly Sins of Contemporary Quantitative Political Analysis”:

The typical paper I receive has some subset – and often as not, the population – of the following irritating characteristics:

- A dozen or so correlated independent variables in a linear model;
- A new whiz-bang and massively complex statistical technique (conveniently available in Stata or R) that is at best completely unnecessary for the problem at hand, since a simple t-test or ANOVA would be quite adequate to extract the few believable results in the data, and not infrequently the technique is completely inappropriate given the characteristics of the data and/or theory;
- Analyzes a data set that has been analyzed a thousand or more times before;
- Is 35 or minus 5 pages in length, despite producing results that could easily be conveyed in ten or fewer pages (as one finds in the natural sciences)

At least now, having taken a quantitative methods course, I now understand why the analysis is useless, where as before I would just skip it to read the conclusions. Actually, I still do this, because as he states, the paper is usually 15+ pages too long. However, I don’t understand what his problem is with Stata.

Advertisements