A Gaussian process (GP) specifies a distribution for values over a set (typically R n ) which could be an interval of time or a region of space, but could also be more arbitrary, such as the space of a set of explanatory variables. It is often used for modeling quantities that are spatially or temporally correlated, such as rainfall. In this way it differs from standard statistical models where the data are assumed to be independent given the model; here the data are assumed to be correlated and the correlation structure is explicitly part of the model.