Standard Errors and Significance Tests for Populations

• In Event: 3269 - Statistical Advances

Abstract

The paper considers an old statistical controversy: is it meaningful to compute and interpret standard errors, significance tests and confidence limits for populations? The paper finds fault with defenses of this practice found in statistical literature, and offers an alternative rational for it. The use of conventional inferential statistical procedures - and limits to their use - are illustrated with an analysis of aggregated data for the age-arrest relationship in the United States.

Author

• David F. Greenberg, New York University