Multinomial smoothing and investigating elevated disease incidence by penalized likelihood.
Date
1997
Authors
Downer, Robert Glen.
Journal Title
Journal ISSN
Volume Title
Publisher
Dalhousie University
Abstract
Description
A compromise is needed between the presentation of highly variable estimates of incidence for subregions and an overall estimate for the region as a whole. Smoothed estimates are obtained by adding an inter-site distance penalty to a constrained multinomial likelihood. An approximation is developed for the smoothed estimate and approximate moments are obtained.
A test statistic is developed based on the smoothed estimates for detecting elevated incidence at a site relative to the others. An appropriate critical region for the test statistic is established. Clusters are defined to be contiguous elevated rates and the technique is evaluated through simulation. Significant increases in detection ability are observed in many situations and the size of the test is controlled. Techniques are also discussed for determining the smoothing parameter from the data.
The overall method is applied to gastric cancer incidence in Nova Scotia, Canada.
Thesis (Ph.D.)--Dalhousie University (Canada), 1997.
A test statistic is developed based on the smoothed estimates for detecting elevated incidence at a site relative to the others. An appropriate critical region for the test statistic is established. Clusters are defined to be contiguous elevated rates and the technique is evaluated through simulation. Significant increases in detection ability are observed in many situations and the size of the test is controlled. Techniques are also discussed for determining the smoothing parameter from the data.
The overall method is applied to gastric cancer incidence in Nova Scotia, Canada.
Thesis (Ph.D.)--Dalhousie University (Canada), 1997.
Keywords
Biology, Biostatistics., Statistics., Health Sciences, Public Health.