Bayesian Analysis of Autism and Mercury Emissions
Author Information
Author(s): Scott M. Bartell, Thomas A. Lewandowski
Primary Institution: University of California, Irvine
Hypothesis
How does administrative censoring affect the ecological analysis of autism in relation to mercury emissions?
Conclusion
The study found that previously reported associations between mercury emissions and autism may have been overestimated due to inadequate statistical methods for handling censored data.
Supporting Evidence
- The Bayesian approach yielded a relative risk estimate of 1.42 per 1000 lbs of air mercury emissions.
- Naive zero-substitution methods produced an inflated relative risk estimate of 4.44 per 1000 lbs of air mercury emissions.
- Censoring of low autism counts was prevalent, affecting 35% of the districts analyzed.
Takeaway
This study shows that when counting autism cases, some numbers are hidden to protect privacy, which can lead to wrong conclusions about how mercury affects autism.
Methodology
The study used a Bayesian censored random effects Poisson model to analyze the relationship between mercury emissions and autism counts.
Potential Biases
The use of naive substitution methods for censored data can introduce bias and underestimate uncertainty.
Limitations
The ecological nature of the data limits the ability to generalize findings to individual-level effects.
Participant Demographics
The study focused on school districts in Texas, with autism counts reported for various districts.
Statistical Information
P-Value
1.18
Confidence Interval
1.07, 1.32
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website