make.hmm.pars.from.Et: Makes 2-state hidden Markov model list suitable for passing...

In david-borchers/hmltm: Line transect estimation with hidden Markov availability models

Makes 2-state hidden Markov model list suitable for passing to est.hmltm via argument hmm.pars.

Description

make.hmm.pars.from.Et creates 2-state hidden Markov availability model specification from mean times unavailable (Eu) and available (Ea).

Usage

make.hmm.pars.from.Et(Ea, Eu, seEa, seEu, covEt = 0, pm = NULL)

Arguments

Ea	vector of length m>=1 specifying mean distance (=mean time * observer speed) animals are in the more available state in one cycle (e.g. dive cycle: surface-dive).
Eu	vector of length m>=1 specifying mean distance (=mean time * observer speed) animals are in the more UNavailable state in one cycle.
seEa	standard error of Ea.
seEu	standard error of Eu.
covEt	vector of length m>=1 containing covariance of each pair of Ea and Eu.
pm	is 2xm matrix containing state-dependent Bernoulli distribution parameters for the m pairs of Ea and Eu, with first being probability of being available when in state i (i=1,2), where i=1 is the more UNavailable state and i=2 is the more available state.

Details

Calculates 2-state hidden Markov model parameters such that the Markov process is in states 1 (more unavailable) and 2 (more available) for mean times Ea and Eu, with Bernoulli state-dependent response probability pm[1,] and pm[2,], respectively. Also constructs a covariance matrix for Ea and Eu. If pm[1,i]=0 and pm[2,i]=1 for animal i, the availability process is a Markov process and the states are actual unavailbile (state 1) and availabile (state 2).

@examples # Some arbitrary numbers for illustration: Ea=c(10,12);Eu=c(100,120);seEa=c(2,3);seEu=c(20,30);covEt=c(10,12) make.hmm.pars.from.Et(Ea[1],Eu[1],seEa[1],seEu[1],covEt[1]) # single animal make.hmm.pars.from.Et(Ea,Eu,seEa,seEu,covEt) # two animals

# Here's how the data porpoise.hmm.pars was created from numbers in Westgate et al. (1995): ppn=c(40,52,36,34,49,60,33)/100 # proportion of time available ET=c(76,44,52,64,70,46,103) # mean dive cycle duration seET=c(48,37,52,65,59,32,67) # SE of mean dive cycle duration cvET=seET/ET # CV of mean dive cycle duration Ea=ET*ppn # mean time available Eu=ET*(1-ppn) # mean time UNavailable # For lack of better info, assume independence of Ea and Eu, and that cv(Ea)=cv(Eu)=cv, # which means that cv=sqrt((cvET*ET)^2/(Ea^2+Eu^2)) # and hence: seEa=Ea*cv seEu=Eu*cv covEt=seET*0 # assume independence porpoise.hmm.pars=make.hmm.pars.from.Et(Ea,Eu,seEa,seEu,covEt)

# Here's how the dataset beaked.hmm.pars were created: Ea=121.824 Eu=1580.256 seEa=9.618659 seEu=134.9212 beaked.hmm.pars=make.hmm.pars.from.Et(Ea,Eu,seEa,seEu)

References

Westgate, A. J., Read, A. J., Berggren, P., Koopman, H. N., and Gaskin, D. E. 1995. Diving behaviour of harbour porpoises, Phocoena phocoena. Canadian Journal of Fisheries and Aquatic Sciences 52, 1064-1073.

david-borchers/hmltm documentation built on Oct. 29, 2023, 9:07 p.m.