poisson_ia.jl

# Utilities for Gibbs sampling with Poisson inter-arrivals


function update_poisson_interarrival_param!(lambda::Vector{Float64},PP::Vector{Int},T::Vector{Int},n::Int,params::Vector{Float64})
    """
    - `lambda`: current Poisson parameter for the interarrival time distribution (to be updated)
    - `PP`: partition/degree vector
    - `T`: arrival times vector
    - `n`: number of observations
    - `params`: parameters of prior Gamma distribution
    """
    K = size(T,1)
    update_poisson_interarrival_param!(lambda,T[end],K,n,params)

end

function update_poisson_interarrival_param!(lambda::Vector{Float64},T_K::Int,K::Int,n::Int,params::Vector{Float64})
    """
    - `lambda`: current Poisson parameter for the interarrival time distribution (to be updated)
    - `K`: number of blocks in partition (=# of arrivals)
    - `n`: number of observations
    - `params`: parameters of prior Gamma distribution
    """
    # truncate support (Julia's truncated Poisson distribution throws an error)
    supp = 0:ceil(100*lambda[1])
    logp = logpdf.(Poisson(lambda[1]),n - T_K .+ supp)
    # sample pseudo arrival K+1 (given λ; doesn't affect distribution of other arrival times)
    T_Kp1 = n + 1 + wsample(supp,log_sum_exp_weights(logp))
    # sample conjugate lambda
    lambda[1] = rand(Gamma(T_Kp1-K-1+params[1],params[2]/(params[2]*K + 1)))
end