Copyright 2009 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2009
We agree with Braun and Kalkbrenner that caution is needed in interpreting our study given that it is based on county-level data.1 However, we are not guilty of either of their criticisms.
Their first criticism is that we inadequately control for differences in urbanicity across counties. We used 3 different statistical approaches to test whether precipitation was positively associated with autism prevalence rates, including a county fixed-effects (within county) specification. The fixed-effects regressions test showed that county-level birth cohorts exposed to high levels of precipitation, relative to the county mean, when younger than 3 years had higher autism prevalence rates relative to the county mean. That we find statistically significant results in these regressions indicates that our findings are not due to differences in urbanicity across counties, because urbanicity did not change much within counties during this period, and any changes are unlikely to be correlated with precipitation.
Waldman M, Nicholson S. Autism Prevalence and Precipitation: The Potential for Cross-Level Bias—Reply. Arch Pediatr Adolesc Med. 2009;163(5):492-493. doi:10.1001/archpediatrics.2009.84