Examples
Getting started expands on the Getting Started section on the homepage and illustrates the use of cmdstanpy to draw samples of Gaussian processes using Stan. The example considers centered and non-centered Parameterizations for Fourier Methods.
Logistic regression illustrates the use of Fourier methods to fit a Gaussian process to binary outcome data.
Effect and importance of padding explores the effect and importance of padding for Fourier Methods when the process does not have periodic boundary conditions.
Gaussian process Poisson regression uses a latent GP with Poisson likelihood for count data. The example considers three different approaches (standard approach using the full covariance matrix, Sparse Approximation, and Fourier Methods) with both centered and non-centered Parameterizations for a total of six models.
Passengers on the London Underground network uses the Sparse Approximation to explore daily passenger numbers on the London Underground network. It illustrates how to combine fixed effects with a latent GP to infer a smooth function that accounts for residual passenger number variability after accounting for the number of interchanges and transport zone of each station on the network.
Density of T. panamensis on a 50 ha plot in Panama uses Fourier Methods to model the density of trees on a 50 hectar plot in Panama. The example highlights the importance of padding the data to attenuate the effects of periodic boundary conditions inherent to the discrete Fourier transform.
Stan introduction provides a brief introduction to Stan using a linear regression example for illustration.