Paul D. Allison "Event History Analysis: Regression for Longitudinal Event Data"

Event History Analysis: Regression for Longitudinal Event Data (Quantitative Applications in the Social Sciences)

後半がむずかった。
特にcompeting risksの話と、最尤法による推定の話はまだまだ勉強が必要。

random disturbanceを入れないことに関して。

Notice also that none of these models has a random disturbance term. They are not deterministic models, however, because there is random variation in the relationship between unobservable dependent variable h(t) and the observed length of the time interval.
[25]

Coxの革新性。

The [Cox's]method relies on the fact that the model can be factored into two parts: One factor contains information only about b1, b2; the other factor contains information about b1, b2, and the function a(t). Partial likelihood simply discards the second factor and treats the first factor as though it were an ordinary likelihood function. This first factor depends only on the order in which events occur, not on the exact times of occurrence.
[34]