Estimators of Seasonal Occurrence Intensity
Author Information
Author(s): Brookhart M Alan, Rothman Kenneth J
Primary Institution: Brigham and Women's Hospital & Harvard Medical School
Hypothesis
Can new estimators provide less bias in estimating seasonal intensity compared to traditional methods?
Conclusion
The new estimator proposed in this study yields less bias and better mean squared error than traditional methods for estimating seasonal intensity.
Supporting Evidence
- Edwards's estimator is biased for small numbers of events.
- The new estimator has less finite sample bias and better mean squared error.
- MLE and weighted least squares are optimal for large studies with strong seasonality.
- Monte Carlo simulations were used to compare estimator performance.
- Confidence intervals for the new estimator were evaluated.
Takeaway
This study looks at how to better estimate seasonal patterns in rare events, like certain types of leukemia, and finds a new method that works better than older ones.
Methodology
The study compares various estimators through Monte Carlo simulations and re-analysis of data on monocytic leukemia seasonality.
Potential Biases
The traditional estimator by Edwards is shown to be substantially biased, especially with small event counts.
Limitations
The estimators assume a single cyclical effect and may not perform well with complex data.
Participant Demographics
Data on monocytic leukemia incidence in England and Wales from 1974–1998.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website