You are not currently logged in.
Access JSTOR through your library or other institution:
Estimation With Inadequate Information
E. L. Lehmann
Journal of the American Statistical Association
Vol. 78, No. 383 (Sep., 1983), pp. 624-627
Stable URL: http://www.jstor.org/stable/2288130
Page Count: 4
Preview not available
As a possible explanation for the existence of absurd UMVU estimators, a class of estimation problems is defined: estimating the probability of an "unobservable" event, for example, estimating the probability of an event of interest occurring over a long period of time from observations over a much shorter period. It is pointed out that in typical cases of this kind, no reasonable unbiased estimator can exist. Consideration of the maximum likelihood estimator suggests the possibility that in fact there may then often exist no reasonable estimators.
Journal of the American Statistical Association © 1983 American Statistical Association