On a compartmental analysis result (Q1906346)

\textit{M. J. Faddy} [Adv. Appl. Probab. 9, 448-461 (1977; Zbl 0379.92011)] seems among the first to formally state the following fundamental result in stochastic compartmental analysis: Let there be \(m\) compartements whose integral states are denoted by \(X_1 (t), \dots, X_m (t)\), into which ``new'' particles may immigrate after \(t = 0\) and then move independently of one another from compartment to compartment and from each compartment to the environment (compartment 0). Assume that the immigration stream consists of independent Poisson processes with the process into compartment \(i\) having intensity function \(b_i (t)\), and that \(X_1 (0) = \cdots = X_m (0) = 0\). For all \(t > 0\), the \(m\) random variables \(X_1 (t), \dots, X_m (t)\) are independent Poisson variables with \[ E \bigl[ X_j (t) \bigr] = \sum_{ k = 1}^m \int^t_0 b_k (u) p_{j,k} (u,t) du, \] where \(p_{j,k} (u,t)\) is the probability that a particle in compartment \(k\) at time \(u\) is in compartment \(j\) at time \(t\) \((t \geq u)\). In addition to making this observation, Faddy provided ingenious yet informal arguments to justify this result and commented, ``although this result is intuitively appealing, the argument that has been used to derive it will undoubtedly be considered heuristic.'' It is of historical interest to note that \textit{J. F. C. Kingman} [J. Appl. Probab. 6, 1-18 (1969; Zbl 0177.21807)] made a similar observation some 8 years earlier in the context of Markov population processes and in effect validated the aforementioned result for the case where all \(b_i(t)\) are constant functions. By transforming the time scale, one can simply generalize Kingman's proof to the result in question. However, an elementary proof in the same spirit but less abstract compartmental analysis language seems more accessible and illuminating.