Censoring assume now that all lives are not observed

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: ule 6: Parametric models: Binomial and Poisson models Binomial model With censoring Assume now that All lives are not observed over the complete year x to x + 1, that is, there may be other decrements other than death (right censoring); and also the possibility of Increments (left censoring) Hence, we observe a life from age x + ai to x + bi (0 ≤ ai < bi ≤ 1) For each life we know di , ai and bi Consider the random variable Di = 0 if the i th life survives from age x + ai to x + bi , and 1 if the i th life dies so that Pr [Di = 1] = bi −ai qx +ai Pr [Di = 0] = 1 −bi −ai qx +ai 6/30 Actuarial Statistics – Module 6: Parametric models: Binomial and Poisson models Binomial model With censoring Maximum Likelihood Estimation Assuming independence, the likelihood of the total sample will be N (bi −ai qx +ai )di (1 −bi −ai qx +ai )1−di L q; d = i =1 where q = (b1 −a1 qx +a1 ,b2 −a2 qx +a2 , · · · ,bN −aN qx +aN ) d = (d1 , d2 , · · · , dN ). Note: unless we are using a (continuous) parametric model for the probabilites of death, there are potentially as many probabilities of death to estimate than data points another possibility is to reformulate all as a function of qx using assumptions (0 ≤ t ≤ 1): 7/30 uniform distribution of deaths: t qx = t · qx , or Balducci assumption: 1−t qx +t = (1 − t ) · qx , or constant force of mortality: t qx = 1 − e −µt Actuarial Statistics – Module 6: Parametric models: Binomial and Poisson model...
View Full Document

This document was uploaded on 04/03/2014.

Ask a homework question - tutors are online