4660 actuarial statistics module 5 parametric methods

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ikelihood estimators In a time-homogeneous Markov jump process with state space {1, 2, · · · , m}, we observe Total number of samples: N (l ) (l ) Ti = N Ti : total waiting time in state i , where Ti l =1 the waiting time in state i of l th life. Nij = (l ) N i =1 Nij : is the total number of transitions from state i (l ) to state j , where Nij is the number of transitions from state i to state j made by l th life. 46/60 Actuarial Statistics – Module 5: Parametric methods: Markov Model General Markov Model Maximum likelihood estimators As usual, we use lower case symbols for the observed samples For life l , the likelihood is m i =1 n (l ) e −µi ti j =i (l ) µijij The total likelihood is (by multiplying over all l ) : N l =1 m i =1 47/60 m i =1 e −µi ti (l ) e −µi ti n j =i µijij n (l ) ij j =i µij = Actuarial Statistics – Module 5: Parametric methods: Markov Model General Markov Model Example (continued) As usual, we use lower case symbols for the observed samples. For life i , the likelihood is proportional to e −(µ+σ)vi e −(ν +ρ)wi µdi ν ui σ si ρri The total likelihood: L (σ, ρ, µ, ν ) N e −(µ+σ)vi e −(ν +ρ)wi µdi ν ui σ si ρri = i =1 −(µ+σ )v −(ν +ρ)w d u s r =e = e e −µv d µ µνσρ e −σ v σs e −ν w ν u e −ρw ρr So the log-likelihood: log L = −(µ + σ )v − (ν + ρ)w + d log µ + u log µ + s log σ + r log ρ 48/60 Actuarial Statistics – Module 5: Parametric methods: Markov Model General M...
View Full Document

This document was uploaded on 04/03/2014.

Ask a homework question - tutors are online