Multiple Comparison of Change Trends in Cancer Mortality/Incidence Rates Taking with Overlapping Regions and Time-Periods

Summary

When analyzing trends in cancer rates, it is common to rely on the so-called Annual Percent Change (APC). For dealing with such a measure of trend, directly age-adjusted rates are usually considered. Classical methods such as pooled t-tests are often applied for comparing APCs of two groups of individuals in a simple way under independence assumption. In practice, it is quite common to find groups of interest for which the independence assumption fails because their regions or periods of time overlap. No one of the papers that deal this problem consider the case where more than two APCs are compared. In this work we propose a Wald-type test-statistic which is not difficult to compute once we provide the estimators of two or more APCs to be compared. These estimators are the minimum power divergence estimators that cover as special case those obtained by maximum likelihood.

This work is related to the stay of Nirian Martín in Harvard University (September 2008 - August 2009), supported by the Real Colegio Complutense. She met Marisa just the first day after arriving in Spain, September 1. In the last conversation Marisa told Nirian that both will meet soon again. Due to her optimistic character it was impossible to imagine they both could not meet again. This paper has been written in memory to Marisa.