select_diffusion_parameters: Parameter selection for the diffusion term

In citiususc/voila: Variational Inference for Langevin Equations

Description

The selection of the amplitude parameters for gaussian process modelling the diffusion term is complicated since we have to link the covariance matrix of a log-normal random variable with our prior belief about the variance of the diffusion term (here, variance may be interpreted as a reasonable estimate for the squared-amplitude of the diffusion). This function helps with the selection of those parameters.

Usage

select_diffusion_parameters(x, samplingPeriod, priorOnSd,
  responseVariableIndex = 1, varX = NULL)

Arguments

x	A time series that may be modelled with a Langevin equation.
samplingPeriod	the sampling period of the time series
priorOnSd	An estimate for the maximum change in amplitude of the diffusion term.
responseVariableIndex	Integer indicating in which of the input dimensions the inference will be made.
varX	The selection of the hyperparametes is based on the variance of the differentiated time series. Although it is estimated by the procedure, the user may specify a different value using this parameter.

Value

A list with a mean (v) and a covariance amplitude for the log-gaussian process (kernelAmplitude).

See Also

See sde_kernel for a very simple example.

citiususc/voila documentation built on May 13, 2019, 7:30 p.m.