fit.gravity: Fit gravity model to movement matrix

In gilesjohnr/hmob: Tools for analyzing human mobility data

Description

This function fits gravity model parameters to a supplied movement matrix using Bayesian MCMC inference. The function defines the model and serves as a wrapper for the run.jags function in the runjags package. Gravity model formula:

θ * ( N_i^ω_1 N_j^ω_2 / f(d_ij) )

Usage

fit.gravity(
  M,
  D,
  N = NULL,
  N.orig = NULL,
  N.dest = NULL,
  n.chain = 2,
  n.burn = 1000,
  n.samp = 1000,
  n.thin = 1,
  prior = NULL,
  DIC = FALSE,
  parallel = FALSE
)

Arguments

M	named matrix of trip counts among all ij location pairs
D	named matrix of distances among all ij location pairs
N	named vector of population sizes for all locations (either N or both N.orig and N.dest must be supplied)
N.orig	named vector of population sizes for each origin
N.dest	named vector of population sizes for each destination.
n.chain	number of MCMC sampling chains
n.burn	number of iterations to discard before sampling of chains begins (burn in)
n.samp	number of iterations to sample each chain
n.thin	interval to thin samples
DIC	logical indicating whether or not to calculate the Deviance Information Criterion (DIC) (default = FALSE)
parallel	logical indicating whether or not to run MCMC chains in parallel or sequentially (default = FALSE)

Value

a runjags model object containing fitted gravity model paramters

Author(s)

John Giles

See Also

Other model: fit.jags(), fit.prob.travel(), rjags.combine(), rjags.mcmc.list(), summarize.hmob()

Other gravity: fit.prob.travel(), sim.gravity.duration(), sim.gravity()

Examples

M <- load.obj(2, './data/movement_data_arrays_full.Rdata')            # M.route
D <- load.obj(1, './data/distance_matrix_named.rdata')                # distance.matrix
N <- load.obj(1, './data/N_pop.rdata')                                # N.pop

M <- M[1:5, 1:5]

districts <- attributes(M)$dimnames$origin 
D <- D[districts, districts]
N <- N[names(N) %in% districts]

out <- fit.gravity(M=M,
                   D=D,
                   N=N,
                   n.chain=2,
                   n.adapt=1000,
                   n.burn=4000,
                   n.samp=4000,
                   n.thin=1,
                   parallel=F)

summary(out)

gilesjohnr/hmob documentation built on Aug. 8, 2020, 1:26 a.m.