dsdive.impute: Impute a complete dive trajectory from partial observations

In jmhewitt/dsdive: Discrete Space Models for Dive Trajectories of Animals

Description

Uses properties of homogeneous Continuous time Markov Chains (CTMCs) to impute trajectory segments between observations. The sampler begins by sampling the number of transitions between observations N, then samples a length N path that connects the trajectory segment at its start and end states.

Usage

dsdive.impute(
  depths,
  times,
  t.stages,
  rate.unif,
  P.raw,
  P.tx,
  n.bins = nrow(depths),
  max.tx
)

Arguments

depths	Depth bin indices visited
times	Times at which each of depths was visited
t.stages	Stage transition times for the dive; will be used to compute the dive stage for each observation
rate.unif	uniformization rate, for standardizing transition rates between states
P.raw	list of continuous time probability transition matrices, and components.
P.tx	list of discrete time probability transition matrices
n.bins	number of rows in the depths matrix
max.tx	maximum number of transitions between observations that will be allowed during imputation

Examples

data('dive.sim')
attach(dive.sim)
attach(dive.sim$params)

# true stage transition times for the dive
t.stages = sim$times[c(FALSE, diff(sim$stages)==1)]

# uniformized transition rate
rate.unif = max(outer(lambda, 2*depth.bins[,2], '/'))

# time between observations
tstep = diff(sim.obs$times[1:2])

# probability transition matrix for observations
P.raw = lapply(1:3, function(s) {
  dsdive.obstx.matrix(depth.bins = depth.bins, beta = beta, 
                      lambda = lambda, s0 = s, tstep = tstep, 
                      include.raw = TRUE, delta = 1e-10)
})

# probability transition matrix for uniformized DTMC
P.tx = lapply(1:3, function(s) {
  dsdive.tx.matrix.uniformized(depth.bins = depth.bins, beta = beta, 
                               lambda = lambda, s0 = s, 
                               rate.uniformized = rate.unif)
})

# impute dive trajectory
x = dsdive.impute(
  depths = sim.obs$depths, times = sim.obs$times, t.stages = t.stages, 
  rate.unif = rate.unif, P.raw = P.raw, P.tx = P.tx, n.bins = nrow(depth.bins), 
  max.tx = 100)

# overlay trajectory over observations
pl = plot(x = sim.obs, depth.bins = depth.bins, stages = sim.obs$stages, 
          errorbars = TRUE, imputed.list = x)

detach(dive.sim$params)
detach(dive.sim)

jmhewitt/dsdive documentation built on May 29, 2020, 5:18 p.m.