emHMM: EM-algorithm to fit a hidden Markov model representing the...
In MarieAugerMethe/CCRWvsLW: Searching strategy movement models

Description Usage Arguments Details References Examples

emHMM finds the maximum likelihood estimate for the parameters of the CCRW by fitting the hidden Markov model through an Expectation Maximization (EM) algorithm.

1 2	emHMM(SL, TA, missL, SLmin, lambda, gamm, delta=c(0.5,0.5), kapp, notMisLoc, maxiter=10000, tol=1e-5)

`SL`	numeric vector containing the step lengths
`TA`	numeric vector containing the turning angles
`missL`	integer vector containing the number of time steps between two steps. If no missing location it will be 1.
`SLmin`	one numeric value representing the minimum step length
`lambda`	numeric vector of length 2 containing the starting value for the lambdas of the two behaviors
`gamm`	2x2 matrix containing the starting value for the transition probability matrix
`delta`	numeric vector value for the probability of starting in each of the two behaviors, default value c(0.5,0.5), which means that you have an equal chance of starting in each behavior
`kapp`	one numeric value representing the starting value for the kappa of the von Mises distribution describing the extensive search behavior
`notMisLoc`	integer vector containing the index of the locations that are not missing
`maxiter`	one integer value representing the maximum number of iterations the EM algorithm will go through. Default = 10000.
`tol`	double: value that indicates the maximum allowed difference between the parameters.

Will return the parameter estimates and the minimum negative log likelihood.

Please refer to Auger-Methe, M., A.E. Derocher, M.J. Plank, E.A. Codling, M.A. Lewis (2015-In Press) Differentiating the Levy walk from a composite correlated random walk. Methods in Ecology and Evolution. Preprint available at http://arxiv.org/abs/1406.4355

For more information on the EM-algorithm please refer to Zucchini W. and I.L. MacDonald (2009) Hidden Markov Models for Time Series: An Introduction Using R. Chapman and Hall/CRC

simPath <- simmCCRW(500,0.9,0.9,0.1,0.01,5,1)
formPath <- movFormat(simPath)
emHMM(formPath$SL,formPath$TA, formPath$missL, formPath$SLmin, 
      lambda=c(0.1,0.1), gamm=matrix(c(0.8,0.2,0.2,0.8),nrow=2),
      kapp=10,notMisLoc=formPath$notMisLoc)