INFORMATION CONTENT OF INTERMITTENT REALIZATIONS FROM FINITE MARKOV CHAINS

摘要

In this paper we study the properties of two special types of intermittent realizations and the corresponding estimation procedures. Incidentally, it may be noted that it is sometimes possible to increase the efficiency of the estimate by deliberately resorting to sampling intermittent observations rather than taking successive observations. Also if the original chain is periodic, the large sample properties of maximum likelihood estimates based on the conventional sample of successive observations no longer hold true. This difficulty can be overcome by using a suitably chosen intermittent sample.