Data-Driven Model Building for Life-Course Epidemiology

Abstract Life-course epidemiology is useful for describing and analyzing complex etiological mechanisms disease development, but existing statistical methods are essentially confirmatory, because they rely on a priori model specification. This limits the scope of causal inquiries that can be made, these suited mostly to examine well-known hypotheses do not question our established view health, which could lead confirmation bias. We propose an exploratory alternative. Instead specifying life-course prior data analysis, method infers directly from data. Our proposed extends Peter-Clark (PC) algorithm (named after its authors) discovery, it facilitates including temporal information inferring observational The extended called PC. obtained afterward perused interesting hypotheses. complements classical confirmatory guides researchers in expanding their models new directions. showcase using set encompassing almost 3,000 Danish men followed birth until age 65 years. Using this set, we inferred role socioeconomic health-related factors development depression.