Unformatted text preview: mply θ (does not
depend on y ).
UNM Another one is the Weibull distribution.
f (y |λ, φ) = λφy λ−1 exp(−φy λ ); y , λ, φ > 0
More ﬂexible than exponential.
In particular, its hazard function is
h(y |λ, φ) = λφy λ−1
which is an ”accelerated failure time model”.
It can be shown that for this distribution
log (−log (S (y |λ, φ))) = log (φ) + λlog (y )
A plot of log (y ) values vs. empirical values of
log (−log (S (y |λ, φ))) may suggest if the Weibull dist. is
UNM Non-parametric estimation
Empirical survivor function (without censoring).
number of subjects with time ≥ y
S (y ) =
View Full Document
- Fall '13
- Probability, NJ, Survival analysis