| Three separate, but related, manuscripts comprise this dissertation on topics in the design and analysis of observational studies. The first chapter develops a new study design called entire matching, which answers the question of how many controls should be matched to a treated subject; in the chapter, entire matching is applied to an observational study of the sentinel lymph node biopsy for the treatment of melanoma. The second chapter is closely related to the first, and incorporates recent developments in the literature, namely the theory behind matching with fine balance without needing to match exactly on levels of a covariate; the question of how many controls is reformulated when fine balance allows the researcher to exert tighter control over certain covariates in an observational study. The third chapter illustrates the power of design to address concerns about unobserved biases in matched studies; the observational study of the sentinel lymph node biopsy is revisited, in which concerns about the limits of matching are mitigated by a clever use of multiple control groups. |
