Characterizations of Competing Risks in Terms of Independent‐Risks Proxy Models (Q4956065)

A survival model \((C,T)\) is considered, where \(T\) is a failure time and \(C\) is a subset of \(\{1,\dots,p\}\) causes of failure. The question is if this model can be reduced to \((T^*,C^*)\), where \[ T^*=\min(R_c,S_r\;:\;c\subseteq\{1,\dots,p\},\;r=1,\dots,r\}), \] where \(R_c, S_p\) are independent, \(R_c\) are absolutely continuous, \(S_r\) are discrete with jumps at \(0<\tau_1<\tau_2<\dots\), and \(C^*\) is the set at which the minimum is attained. The model \((T^*,C^*)\) is called an independent-risks proxy model. The author derives necessary and sufficient conditions for \((T^*,C^*)\) existence: For all \(c\subseteq\{1,\dots,p\}\) and all \(\tau_l\), \[ \Pr(C=c,T=\tau_l)/\Pr(T>\tau_l)= \prod_{j\in c}\{\Pr(C=j,T=\tau_l)/\Pr(T>\tau_l)\}. \] Some examples are considered.