Detecting tornadoes : a spurious population effect?.

Includes bibliographical references

Uses National Severe Storms Forecast Center's information on 31,969 tornado segments occurring in the continental U.S. during the 1950-1990 period, and ecological information on all counties experiencing tornadoes during this period available from the U.S. Census to model the occurrence of weak tornadoes which are most likely to go unreported. The relative complexity of the demographic pattern of counties is insignificantly related to the proportion of counties with weak storms. Metropolitan and other urban counties do not have higher odds of weak tornadoes than rural counties. Inferentially, these results fail to support the prevailing interpretation in the scientific literature that the known relationship between higher frequency of tornado occurrence and demographic complexity is spurious. An alternative interpretation is suggested