Epidemiological Study Designs and Their Statistical Implications

This chapter introduces the methodological foundations of epidemiological research, emphasizing how the choice of study design determines the research questions that can be answered, the structure of the data collected, and the statistical methods that can be appropriately applied. Students learn to distinguish descriptive studies, which characterize disease frequency and distribution according to person, place, and time, from analytical studies, which investigate associations between exposures and health outcomes. The chapter discusses the principal epidemiological designs, including cross-sectional, cohort, case-control, ecological, and randomized controlled trials (RCTs). Each design has specific strengths, limitations, and appropriate measures of disease frequency or association. For example, cross-sectional studies are particularly useful for estimating prevalence, cohort studies allow the estimation of incidence and risk ratios, case-control studies primarily use odds ratios, and RCTs provide a strong framework for evaluating causal effects of interventions. These concepts help Statistics students understand how research objectives and study design influence the validity and interpretation of epidemiological evidence.

A second major component focuses on the statistical consequences of different study designs, particularly the selection of appropriate estimands, measures of association, and analytical models. Students explore prevalence ratios, risk ratios (RR), risk differences (RD), odds ratios (OR), incidence rate ratios (IRR), and hazard ratios (HR), together with confidence intervals and hypothesis testing. The chapter introduces regression methods for controlling confounding, including logistic regression, Poisson regression, and Cox proportional hazards models. It also examines the special characteristics of ecological studies, especially the ecological fallacy, which occurs when population-level associations are incorrectly interpreted as individual-level relationships. In addition, students learn how to evaluate internal validity, which concerns the credibility of findings within the study population, and external validity, which concerns the generalizability of results to a target population. Important threats to validity—including selection bias, information bias, confounding, recall bias, and loss to follow-up—are discussed alongside strategies to minimize their effects. Through these topics, students develop an understanding of why statistical significance alone is insufficient to establish a valid association or causal relationship.

The final component of the chapter addresses sampling design, sample size determination, and the implications of sampling for statistical inference. Students examine the differences between target populations, source populations, and sampling frames, as well as probability sampling methods such as simple random, stratified, cluster, and multistage sampling. The material explains how sample size depends on the significance level, statistical power, expected effect size, and population variability. It also discusses design effects, sampling weights, nonresponse, missing data, and analytical approaches such as inverse probability weighting, multiple imputation, generalized estimating equations, and mixed models. These concepts demonstrate why complex sampling structures must be incorporated into statistical analyses to obtain appropriate standard errors, confidence intervals, and population estimates. Overall, the chapter equips Statistics students with the methodological and analytical skills needed to select suitable epidemiological study designs, estimate health-related parameters, assess bias and validity, and produce reliable evidence for health research and public health decision-making.