scsgmm 0.1.0

Sequential Conditional Sampling from GMMs

scsgmm

Sequential Conditional Sampling from Gaussian Mixed Models (SCS-GMM) is a method to fit simple parametric models to univariate and multivariate time series and generate synthetic realisations. The method was inspired by Sharma et al (1997) Streamflow simulation: A nonparametric approach [https://doi.org/10.1029/96WR02839], and originally implemented using KDEs in the package scskde. The next logical step was to try an analogous parametric approach, the simplest of which is to replace the KDEs with GMMs. The approach supports

• lag orders >= 1
• arbitrary seasonality
• vector-valued processes
• exogenous forcing
• arbitrary dependence structure