networkTomography (version 0.3)

move_step: Move step of sample-resample-move algorithm for multilevel state-space model

Description

Function to execute single MCMC-based move step for bayesianDynamicFilter. This can use two types of stopping rules: number of iterations or number of accepted moves for the X particles. The former is used by default, but the latter adapts better to low acceptance rates (sometimes with substantial computational cost). Most updates in this algorithm are Metropolis-Hastings with customized proposals.

Usage

move_step(y, X, tme, lambda, phi, lambdatm1, phitm1, prior, A, A1_inv, A2, rho, tau, m = ncol(X), l = nrow(A1_inv), k = length(lambda), ndraws = 10, minAccepts = 0, verbose = FALSE)

Arguments

y

numeric vector (length l) of observed link loads

X

matrix (m x k) of particles for OD flows, one particle per row, in pivoted order

tme

integer time index currently used in estimation

lambda

matrix (m x k) of particles for OD intensities, one particle per row, in pivoted order

phi

numeric vector (length m) of particles for phi

lambdatm1

lambda matrix (m x k) of particles for OD intensities from previous time, one particle per row, in pivoted order

phitm1

numeric vector (length m) of particles for phi from previous time

prior

list containing priors for hyperparameters; see bayesianDynamicFilter for details

A

routing matrix (l x k) for network

A1_inv

inverse of full-rank portion of routing matrix (l x l)

A2

remainder of routing matrix (l x k-l)

rho

numeric fixed autoregressive parameter for dynamics on lambda; see reference for details

tau

numeric fixed power parameter for variance structure on truncated normal noise; see reference for details

m

integer number of particles

l

integer number of observed link loads

k

integer number of OD flows to infer

ndraws

integer number of draws to perform (can be overriden by minAccepts)

minAccepts

integer minimum number of acceptances before results are returned; activates alternative stopping rule if >= 1

verbose

logical activates verbose diagnostic output

Value

list containing updated values of X, lambda, and phi

References

A.W. Blocker and E.M. Airoldi. Deconvolution of mixing time series on a graph. Proceedings of the Twenty-Seventh Conference Annual Conference on Uncertainty in Artificial Intelligence (UAI-11) 51-60, 2011.

See Also

Other bayesianDynamicModel: bayesianDynamicFilter; buildPrior