Aalen-Johansen Estimator

Definition

The Aalen-Johansen estimator is a multi-state (matrix) version of the Kaplan–Meier estimator for the hazard of a survival process. The estimator can be used to estimate the transition probability matrix of a Markov process with a finite number of states.

Estimator

The position in state space for an entity $i$ in continuous time $t$ is a Random Variable $R^i(t)$ taking values in the state space S (We assume a finite state space $S ={0, \dots ,D}$).

The estimator is given by the expression

$T^{mn}(s,t) = \prod_{k=1}^{K} (I^{mn} + \Delta A^{mn}(t_k))$

$T^{mn}(s, t)$ is the transition matrix element from time s to time t, the mn-th element of the matrix denotes the probability that the Markov process starting in state m at time s will be in state n at time t. The summation is over all times $t_k$ where transition events are observed (a total of K).

Off-diagonal elements

The estimation of the transition intensities $A^{mn}(t_k)$ at any time $t_k$ where transitions are observed is simply by counting:

$\Delta A(t_k)^{mn} = \frac{\Delta N^{mn}(t_k)}{Y^{m}(t_k)}$

where $\Delta N^{mn}(t_k)$ is the number of transitions observed from state m to state n at time $t_k$ and $Y^m(t_k )$ is the number of entities in state m right before time $t_k$

Diagonal elements

The diagonal elements are given by

$\Delta A(t_k)^{nn} = \frac{- \Delta N^{n}(t_k)}{Y^{n}(t_k)}$

where $\Delta N^{n}(t_k)$ is the number of transitions away from state n at time $t_k$

See Also

The Aalen-Johansen estimator is equivalent to the cohort method when the latter is applied to very short intervals