Sample Rotation Method for Non-response
by
V. K. Singh
Department of Statistics, Banaras Hindu University, India
Coauthors: S.S. Mandal (Medical Records Unit, Chittaranjan National Cancer Institute, S.P. Mukhrjee Road, Calcutta, India)

It is a common experience that the required information cannot be collected always on all the units which are sampled; either it be a one-point survey or be a repeated survey. This is particularly true when units are human beings and a mail survey is adopted. This incompleteness of information is termed "non - response". It is a well known fact that such error do occur and occur in spite of all the care that would generally be accepted as sufficient in other situations.

Depending upon the aim of the survey, there are many considerations in repeated surveys as regards the selection of samples at different occasions. Intuition suggests that retention of a part of the sample over occasions may provide efficient sampling strategies. The use of information on an appropriate auxiliary information at the current occasion may also lead to better estimation procedures.

The present paper is aimed at developing efficient classes of estimators in repeated surveys in presence of non - response. The technique of sub - sampling the non - respondents is used to develop these classes. Some important properties of the classes along with particular members of the family of estimators have been discussed. The problem of determining optimum size of the replaced sample is also discussed. The strategies have been compared with other existing strategies for their efficiencies.

Date received: October 30, 2000