The information in aggregate data from Markov chains

From MaRDI portal
Publication:3683376





The authors discuss estimation in Markov chains from aggregate data (i.e. data consisting of mere counts of the total number of individuals in each state at each time point). The purpose of the paper is to investigate the information about various aspects of a chain, and to measure the loss of information incurred by aggregating data rather than observing the complete histories of each individual (panel data). A generalized least squares estimator and its asymptotic properties are first reviewed. Then two-state chains and finally k-state chains are considered and comparisons made with the panel data case. The main general conclusion is that although aggregate data are generally much less informative than panel data about individual transition probabilities, they provide relatively good estimation of equilibrium distributions and of mean occupancy numbers for the states of the chain.




Cited in
(23)








This page was built for publication: The information in aggregate data from Markov chains

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3683376)