R/GetRho.R

In bcpa: Behavioral Change Point Analysis of Animal Movement

#' Characteristic time / auto-correlation for irregular time series
#' 
#' Estimates characteristic time \eqn{\tau} or auto-correlation \eqn{\rho} from 
#' a gappy time series dataset.  It works first by estimating the mean and 
#' standard deviation directly from the time series X, using these to 
#' standardize the time series, and then optimizes for the likelihood of the 
#' value of \eqn{\tau} or \eqn{\rho}. 
#' 
#' @aliases GetRho2
#' @param x  vector of time series values.
#' @param t  vector of times of measurements associated with x.
#' @param tau whether or not to estimate time scale \eqn{\tau} (preferred) or 
#' autocorrelation \eqn{\rho}.
#' @return Returns a vector of length two: the estimate and the negative 
#' log-likelihood
#' @author Eliezer Gurarie

GetRho <- function(x,t, tau = TRUE){
    getL <- function(rho)
      GetL(x, t, rho, tau)
    if(!tau)
      o <- optimize(getL,lower=0,upper=1,tol=.001, maximum=TRUE)
    else
      o <- optimize(getL,lower=0,upper=t[length(t)] - t[1],tol=.001, 
                    maximum=TRUE)
    # MUST REMEMBER: 1) rhohat  2) LL
    return(c(rho.hat = o$maximum, LL = o$objective))
  }

GetRho2 <- function(x,t, tau = TRUE){
    if(length(x)>2){
      getL <- function(rho)
        GetL(x, t, rho, tau)
      if(!tau) o <- optimize(getL,lower=0,upper=1,tol=.001, maximum=TRUE)
      else o <- optimize(getL,lower=0,upper=t[length(t)] - t[1],tol=.001, 
                      maximum=TRUE)
      
      # MUST REMEMBER: 1) rhohat  2) LL
      return(rho.hat = o$maximum)
    } else return(NA)
  }