Mapping bone cancer death rates in Pennsylvania counties

The reliability of different arrangements of a single data set, bone cancer deaths by county in Pennsylvania 1960–1976, is examined to examine the reliability of geographic series before they are used to explore environmental cause hypotheses.

Abstract

Computer graphics of areal death rates permit “eyeballing” spatial variations. Such projections help in finding “hotspots” or broader geographic trends. This paper examines the reliability of different arrangements of a single data set, bone cancer deaths by county in Pennsylvania 1960–1976. For this uncommon disease repeat reliability between two component time periods is low. Furthermore, although correlation between two summary indexes—directly adjusted death rate and years of life lost—is high (r = 0.8), there are notable differences between the lists of areas high for each. Bone cancer deaths in Pennsylvania have no reliable and statistically significant highs or lows; aggregations of countries are also blurry. Using data for a rare disease minimizes reliability in this example. Even so, the reliability of geographic series should be thoroughly examined before they are used to explore environmental cause hypotheses.