In both observational and experimental studies, common methods for estimating treatment effects can be statistically biased if certain assumptions are not met. Moreover, the assumptions required for some methods to be unbiased can be difficulty to satisfy. This dissertation explores two common pitfalls in assessing treatment effects, and provides straightforward solutions to issues with current methods. The first chapter of this dissertation discusses the issues with using regression adjustment to estimate treatment effects for an observational study, focusing on the “weighting problem” that arises in the estimated effect from this method. The second and third chapters of this dissertation move to an experimental setting, in which researchers would like to condition on a post-treatment variable of interest, but cannot as this would lead to a statistically biased estimate. We propose a framework for accounting for post-treatment variables and obtaining a valid range of estimates for the treatment effect among a subgroup of interest. We demonstrate the use of this approach through both simulated and applied examples.