Power Calculations for Micro-Geographic ‘Hot-Spots’: How We Learned to Love the Finite Population Correction

• In Event: Crime and Place: The Baltimore Longitudinal Hot Spots Study: Methodological Challenges and Lessons Learned

Abstract

One component of the Baltimore Longitudinal Hot Spots Study involves surveying multiple individuals within each street segment included in the study. For some analyses, these survey responses will be used to estimate street segment level variables of interest that are not otherwise available. While response rates given contact have generally been high, contact rates have been more problematic. The bad news is that as always, smaller sample sizes mean less precision in these street segment level estimates. The good news is that the impact of reduced sample size is mitigated substantially by taking into account the finite population correction (FPC). The FPC is a powerful factor because we found fewer total households per micro-geographic region than we had anticipated. Moving from a within street segment level sample size of 10 down to 7 results in a standard error of .28σ with the FPC which is lower than the standard error at n = 10 without the FPC.

Authors

• Amelia Haviland, Carnegie Mellon University

• David Weisburd, George Mason University / Hebrew University

• Justin Ready, Arizona State University

• Brian Lawton, George Mason University