13_AS_4a_lec_a

# Linear log time model log x z t w dene x0 exp

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

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: accelerated failure-time models Consider the linear log-time model: log X = µ + β Z T + σ W . Deﬁne X0 = exp(µ + σ W ) and let S0 (x ) be the survival function of X0 . Then the survival function of X : S (x ; Z ) = Pr (X &gt; x ) = Pr exp µ + β Z T + σ W &gt; x = Pr exp [µ + σ W ] &gt; x exp −β Z T = Pr X0 &gt; x exp −β Z T = S0 x exp −β Z T Hence, the linear log-time model is equivalent to the accelerated failure-time model with θ = −β and S0 (x ) = Pr (exp(µ + σ W ) &gt; x ). 6/15 Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples 1 Linear log-time model 2 Accelerated failure-time model 3 Relationship between the log-time and accelerated failure-time models 4 Examples 7/15 Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples A variety of distributions can be used for the lifetime distribution, some common distributions: Exponential Weibull Log normal Log logistic ··· See K&amp;M, Chapter 12. 7/15 Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples Weibull Distribution The Weibull distribution with scale parameter λ, λ &gt; 0, shape parameter α, α &gt; 0, x &gt; 0 has a hazard rate of µ (x ) = λαx (α−1) This is very ﬂexible. It allows: increasing hazard rates if α &gt; 1, decreasing hazard rates if α &lt; 1, constant hazard rates if α = 1 8/15 Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples K&amp;M, page 23 9/15 Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples Accelerated failure-time representation: S (x ; z ) = S0 x exp θZ T where the baseline survival function S0 (x ) = exp (−λx α ) . Equivalent linear log-time representation: Y = log X = µ + β Z T + σ W where W has the standard extreme value distribution exp(w − e w ), 10/15 −∞ &lt; w &lt; ∞. Actuarial Statistics – Module 4a: Parametric models: Introduction to more advanced parametric models Examples Proportionality The model has prop...
View Full Document

## This document was uploaded on 04/03/2014.

Ask a homework question - tutors are online