R/fit_hawkes.r

In cmjt/smoke: Fitting apatial and apatio-temporal models to smoking logs

#' Fitting a hawkes (temporal) point process model
#'
#' Fits a simple univariate point process model
#'
#' @param params a numeric vector of length three of starting values
#' for the paramerers mu, alpha, beta respectively.
#' @param times a numeric vector of times
#' @param method the method used by optim for maximisation by default "L-BFGS-B"
#' @return estimates parameter values
#' 
#' @export

fit.hawkes <- function(params, times,population = NULL, marks = NULL, method = "L-BFGS-B"){
    if(!is.null(marks)){
        fit <- optim(params,loglik.marked.hawkes, times = times, marks = marks,method = method)
        rownames <- c("mu","alpha","beta","nu","delta")
    }else{
        if(!is.null(population)){
            fit<- optim(params,loglik.hawkes.population, times = times, population = population,method = method)
            rownames <- c("mu","alpha","beta")
        }else{
            fit<- optim(params,loglik.hawkes, times = times,method = method)
            rownames <- c("mu","alpha","beta")
        }
    }
    pars <- fit$par
    nrow <- ifelse(!is.null(marks),5,3)
    out <- matrix(NA,ncol = 2, nrow = nrow)
    out[,1] <- pars
    rownames(out) <- rownames
    colnames(out) <- c("param","se")
    out
}