Hidden pools, hidden modes, and visible repeated eigenvalues in compartmental models (Q1820704)

The authors address the problem of compartmental models where the number of compartments is larger than that of the eigenvalues appearing in the output. If some eigenvalues have multiplicity larger than 1, systems where the number of ''visible modes'' is not equal to the number of compartments are also analyzed. Some results attained by \textit{S. Parter} [J. Soc. Industr. Appl. Math. 8, 376-388 (1960; Zbl 0115.248)] and \textit{J. Z. Hearon} [Ann. New York Acad. Sci. 108, 36-68 (1963; Zbl 0144.098)] are revisited in order to find criteria for studying ''visible eigenvalues'', ''visible repeated eigenvalues'' and ''hidden pools'' that are investigated by inspecting the compartmental diagram structure. In particular, properties of strongly connected components and tree components of the graph are exploited and a number of properties are illustrated which are of interest for studying both the system matrix spectrum and its implications on the output function. Particular attention is focussed on catenary and mamillary systems as compartmental model components and properties of a given experiment model configuration are derived by means of its decomposition.