An incomplete data problem arises when sample realizations are not fully observable: some realizations may be entirely or partially missing; some variables may be interval-measured. Whatever the specific form of the incomplete data problem, the generic consequence is imprecise identification of the population distribution generating the data. This paper describes completed and ongoing research showing how incomplete data problems lead to imprecise identification of regressions and of parameters solving extremum problems.
Keywords. identification regions, interval data, missing data, nonparametric regression
Format. PDF
Paper Download
The paper is availabe in the following sites:
Authors addresses:
Department of Economics
Northwestern University
2003 Sheridan Road
Evanston, IL 60208
E-mail addresses:
| Charles Manski | cfmanski@northwestern.edu |