Modern science and engineering models depend on several input parameters. Parameter studies, such as uncertainty quantification or optimization, become unwieldy when the number of parameters is large---particularly if evaluating the model is computationally expensive. The scientist might try to rank the parameters in order of importance to enable efficient methods that exploit some anisotropy in the model. We will examine several popular methods for ranking the relative importance of a model's input parameters including the elementary effects of Morris and the variance-based Sobol indices. We will discuss the difference between local and global uncertainty metrics. And finally, we will discuss subspace-based sensitivity analysis, e.g., active subspaces, which seek important linear combinations of the model's input parameters.