fitExpGP: Decay fit with modulation of mean depth by Gaussian Process

In ppernot/FitOCTlib: Stan-based functions to fit OCT signals

Description

Decay fit with modulation of mean depth by Gaussian Process

Usage

fitExpGP(
  x,
  y,
  uy,
  dataType = 2,
  Nn = 10,
  theta0 = NULL,
  Sigma0 = NULL,
  lambda_rate = 0.1,
  lasso = FALSE,
  method = "sample",
  iter = 50000,
  prior_PD = 0,
  alpha_scale = 0.1,
  rho_scale = 1/Nn,
  gridType = "internal",
  nb_chains = 4,
  nb_warmup = 500,
  nb_iter = 1000,
  verbose = FALSE,
  open_progress = TRUE
)

Arguments

x	a numeric vector
y	a numeric vector of responses
uy	a numeric vector of uncertainty on 'y'
dataType	an numeric (1 or 2) defining the type of data
Nn	number of control points
theta0	theta prior mean vbalues
Sigma0	theta prior corvariance matrix
lambda_rate	scale of ctrl points prior
lasso	flag to use lasso prior
method	choice of optimization method in c('optim','sample')
iter	max. number of iterations for 'optim'
prior_PD	flag to sample from prior pdf only
alpha_scale	SD scale of GP
rho_scale	relative correlation length of GP
gridType	type of controle points grid ('internal' does not contain boundaries)
nb_chains	number of MCMC chains
nb_warmup	number of warmup steps
nb_iter	number of steps

Details

Bayesian inference of the parameters of an exponential model with modulation assuming an uncorrelated normal noise y(x) ~ normal(m(x),uy(x)); m(x) = theta[1] + theta[2]*exp(-dataType*x/theta[3]*(1+l(x)));. l(x) is defined by a GP with fixed positions 'xGP', variance and correlation length.

Value

A list containing

fit

a stanfit object containg the results of the fit

xGP

a vector of coordinates for the control points

method

same as input

prior_PD

same as input

lasso

same as input

Author(s)

Pascal PERNOT

ppernot/FitOCTlib documentation built on April 11, 2020, 1:55 a.m.