Detecting spatial clustering for discrete, censored, or longitudinal outcomes

Spatial clustering detection methodology has become increasingly necessary with the rapid development of GIS technology and the abundance of studies that address the effects of hazardous exposures, such as petrochemicals and air pollution. Often, it is difficult to quantify such exposures on an individual level and therefore alternative techniques, such as spatial clustering detection, are necessary to reveal evidence of spatial relationships of exposure and disease outcome. This thesis focuses on developing spatial clustering detection methodology for three different types of outcomes: binary, censored, and longitudinal.; Chapter 1 investigates spatial clustering for binary outcomes coming from both matched and unmatched study designs. Numerous methods have been previously proposed for detecting spatial clustering when the outcomes are assumed independent. However, few have proposed methods to handle spatial clustering where part of the dependence is due to study design (e.g. matched pair studies). An extension of the spatial scan statistic (Kulldorff, 1997) is proposed for the detection of spatial clustering when the data are collected through a matched case-control mechanism. Further, a robust clustering detection method, specifically, a cumulative geographic residual test is proposed, which allows for discrete outcomes, matched or unmatched. Power comparisons between the spatial scan statistic and cumulative geographic residual test are made via simulations. Utilization of these methods is illustrated by a matched case-control study investigating the impact of petrochemical exposure on childhood brain and leukemia cancers.; Chapter 2 extends methods developed in Chapter 1 for failure time outcomes with censoring. Specifically, an extension of the spatial scan statistic is introduced for data with failure time outcomes using the log rank test statistic. Further, an extension of the cumulative geographic residual method is proposed that utilizes the principle of cumulative martingale residuals for censored outcomes. Application of these methods is illustrated by the Home Allergens and Asthma prospective cohort study looking at the relationship of environmental exposures with asthma, allergic rhinitis/hayfever, and eczema.; The last Chapter presents a method to handle detection of spatial clusters of repeated measured outcomes while accounting for moving. It provides an extension of the cumulative geographic residual utilizing generalized estimating equations (GEE) methodology. It further proposes a time-dependent clustering detection method for to define spatial clusters over time. Application of these methods is illustrated by the Home Allergens and Asthma prospective cohort study, introduced in Chapter 2, studying the relationship between environmental exposures and the repeated measured outcome, persistent wheezing in the last 6 months.