Interactive and dynamic visualization of high-dimensional data

As the dimensionality of data increases, so does the difficulty of comprehensive visualization. Traditional linear projections view discrete pairs of linear components. Tours are a class of linear projections that animate over small changes to the projection. The manual tour uniquely enables an analyst to steer a path of bases. This thesis discusses a package that facilitates their creation. A user study finds the manual tour outperforms traditional visualizations on a variable attribution task. This work introduces a novel approach to examine the support of nonlinear model explanations with the radial tour. An accompanying package facilitates this analysis.