fit_kbr_sde: Kernel Based SDEs Estimation

In citiususc/voila: Variational Inference for Langevin Equations

Description

Estimates the drift and diffusion terms of a Langevin Equation using the Kernel Based Regression (KBR) method from a one-dimensional time series.

Usage

fit_kbr_sde(x, h, kernels = c("normal", "normal"), driftBw = 0.5,
  diffBw = 0.5, nSim = 500, nthreads = 1, solveTiesDrift = c("maxDrop",
  "minArg", "maxArg", "minVal", "maxVal", "na"), solveTiesDiff = c("maxDrop",
  "minArg", "maxArg", "minVal", "maxVal", "na"), driftErrorBw = 0.1,
  diffErrorBw = driftErrorBw, plotErrors = TRUE)

Arguments

x	A univariate vector representing the time series
h	The sampling period of the time series
kernels	A vector of 2 strings specifying which kernels should be used in the KBR estimation for the drift (first component of the vector) and the diffusion (second component). It is currently ignored since only the Gaussian kernel is supported.
driftBw, diffBw	Bandwidth of the Gaussian kernel used to estimate the drift/diffusion term
nSim	= 500 Number of simulations used to calculate the delta-error (see references)
nthreads	= 1 Number of threads to be used during the computation
solveTiesDrift, solveTiesDiff	A string specifying the strategy to be used to break ties between two different bandwidths with the same delta-error for the drift and diffusion terms, respectively.
driftErrorBw, diffErrorBw	Since the delta-errors are noisy, a rolling mean is used before selecting the best bandwidth. These parameters specify the width of the rolling mean (in units of bandwidth) used to smooth the drift-errors and the diffusion-errors, respectively
plotErrors	Boolean value. Plot delta-errors?

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