Teaching statistics next semester, adding in this section, which I am tempted to name the “EJ Antoni Memorial Module”.

Before credible statistical analysis can proceed, one has to be sure that (1) one knows definitions, (2) one knows the attributes of the data, (3) one knows the relative reliability of alternative measures of the same variable.

In conducting basic statistical analysis, it is useful to (1) report diagnostics, (2) remember what units the variables are measured in, (3) what understand how a particular procedure works.

Here are some egregious examples of failures to heed these suggestions, either by error or by intent.

• Know your definitions. If you’re going to write a paper on recessions, you should know what the technical definition is (it’s not 2 consecutive quarters of negative growth).
•  Understand your variable, and what it includes (import deflators), then look it up! The BLS import deflator doesn’t include tariffs.
•  Don’t use a series that is noisier than the other. For instance, don’t use CPS based government employees count instead of CES based government jobs count, if you want a more precise measure.
• Report regression statistics if you’re going to cite regression results.  A dissertation without a single reported R-squared or F-test or DW statistic is eyebrow-raising.
•  Know what units your variables are measured in, including when running regressions. For instance, if the dependent variable is an interest rate measured in percentage points. and the independent variable is an interest rate measured in decimal form, then the coefficient, even if “small” at 0.04, actually quite large when expressed in percentage point for percentage point terms.

Addendum 9/10:

Reproducibility. If you’re going to make a stunning conclusion (like, we’ve been in recession since 2022), then provide the source data to replicate your results (I did it for this paper).