Researchers often assess causal effects of educational programs or policies using educational assessment data. This dissertation explores novel methods of estimating causal effects in educational assessment data and is broken into three parts. The first part proposes a regression discontinuity design with an ordinal running variable to assess the effects of extended time accommodations for the National Assessment of Educational Progress. The second part investigates how to enhance the performance of machine learning methods to estimate causal effects in multilevel observational data. The third part discusses how to estimate effect heterogeneity that arises from unobservable, latent characteristics by using machine-learning -based methods for causal inference. Overall, the methods from each part provide investigators with modern tools to estimate causal effects in increasingly large and complex educational assessment data.