Although randomized, controlled, double blind experiments (RCDBE) are considered the gold standard for evidentiary inference, many instances exist where a RCDBE may be unethical or impractical. Moreover, a RCDBE may not be reflective of real-world settings. Such instances may warrant reliance on data from studies from more practical designs to support decision and policy making—e.g., strengthening comparative effectiveness profile of a therapeutic product or medical device using data from routine clinical practice or disease registries. The design, conduct, and statistical analyses of departures from the RCDBE can and should mirror randomized counterpart. Departures could include non-randomized, non-controlled, non-double blinded, or any combination of these. Participants will learn methods for designing credible real-world clinical studies that mimic traditional randomized trials, and statistical analysis of data therefrom. These methods derive from causal inference framework, including propensity score (matching, inverse probability of treatment weight, stratification by propensity score), genetic matching, etc. Best methodological practices for improving credibility of findings will be discussed. Software implementation of these methods using SAS and R will be demonstrated.