NIH Research Festival
Etiological factors identified in cancer epidemiology studies may be missed or be inconsistent across studies due to variation in etiology across disease subtypes (etiological heterogeneity). These subtypes may be unknown or unmeasured in many population-based studies. We develop new statistical methodology for both identifying and modeling etiological heterogeneity in epidemiologic studies. First, we develop a simulation-based test for etiologic heterogeneity where repeated subsamples of the study data are compared with a reference distribution obtained by simulating under a model without etiologic heterogeneity. In a second approach, we propose a latent class model where etiological subtype is unobserved. Results from simulation studies show that our test controls type 1 error and has reasonable power in some settings. Simulations also show that our proposed modeling strategy works in estimating the effects imposed by the unknown etiological subtype. We apply our methods to a Head and Neck cancer dataset and show that it is able to detect the heterogeneity that is known to exist in this dataset.
Scientific Focus Area: Epidemiology
This page was last updated on Monday, September 25, 2023