Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Applied Statistics Unit (ASU-Kolkata)


Chaudhuri, Arijit (ASU-Kolkata; ISI)

Abstract (Summary of the Work)

A common practical problem to whlch a survey - sampler has frequently to address himself 1s one of sampling a given finite population on successive occasions. One of the relevant issues requiring one's attention then 1s to adopt a suitable sampling strategy to estimate the population total of a variate of interest an the current occasion in an optimal manner. Here one has of necessity to take care to utilize the accumalated data on that variate procured in course of the survey along with other auxdliary information on one or more additional variables that may also incidentally be available, Several such strategies are well known in the literature. In this context significant contributions have been from Jessen (1942), Pattereon (1950), TIkkiwal (1950, 1951, 1964, 1965), Keyfitz (1951), Hansen-Hurwite and Madow (1953), Kulldorff ( 1963), Rao, J:N.K. and Graham (1964), Raj (1965), Ghangurde a nd Pao J.N.K. (1969), Fellegi (1966), Singh (1968), Avadhani and Sukhatme (1970), Ravindra Singn (1972), Sen (1973), Chotai ( 1974) and Lanke (1975). Most of the strategies involve Jelection of units with equal prebabilities on each occasion in on: or more stages. Unly those due . mainly to Raj (1965), Changurde ani Fao J .N.K. (1969) arut Cotai (1974) call for selection with varying, probabilities. Since tihe problem is formulated in a finite population set-up we cennot have a best strategy in general. So, setất we mayain at is to have one that may be claimed to be optimal in a certain senae and to get equipped with tools with Which we may udge thecireumstances in which a proposed one may be anticipated to farv better or worse than its possible campetitors and thus to le: able to choose a course of action to follow in particular s ituations. So, a major 'portion of the thesis 1s devoted to sausting alditional samplire stratiles for estimating fintte population totals on a current o cension, by choocing samples in two or more occasions in one or more atagen with varying probab1litieo of selection using auxillary informatlon in the form of what, are conmonly known as size-measures alorg with knownleue of values of the variate of interest on prevlous occasiona as vell. For the suggested strategles we malnly seck to recanmend proportions of units to be: mat ched on successive occasions ard campare the efficiencies of the resulting strategies relative to their rivals often under varlous models popularly treated in survey sampling 11terature. Incldentally, as we obser": that an appropriate spling strategy to employ on two occasions depends on the appropriateness of the strategies for sampling: separately on the current and the previous occasions, we also pay heed to hơw one should choose an optimal sampling strategy on a single occasien alone. Literature is replete with results on this issue. Yet we f tnd it, worthwhile for thesake of campleteness to extend the study of relati efficiencies of several well-kmown strategies for sampling on a single occasion. Further, we note a few associated problems in the context of sampling on a single occasion, namely, determination of opt imal- sample s ize, removing the bias of product estimators without losing efficiency, non-negative estimation of variance, estimation of variance in two-stage sampling. Results of such queries are presented in the chapters 2 a nd 3. These are followed by our st udy of strategies appl1- cable on two occasions for selecting the samples only in one stage, presented in the chapter 4. The chapter 5 takes care of those strategies applied on successive occasions 11 two stages. In chapter 6 we have proposed same strategics f:r sarling 1 1e then tw oecsions. In the final chapter we take up another spécific problem of uni-stage sarpling on two occasicns. Here some aspects of the sampling schemes to be adopted for the two occasions are supposed to be fixed and subject to the c onsequential ccnstraints on them, the problem we treat s about meximization of the expected rumber of cormon units to be surveyed on Loth the cccasions - a step obviously to relieve the pressure on the Luciget. We offer a few alternatives to the earlier results in this area due to Keyfitz (1951), Fellegi (1966) and Lanke (1977) and make a few comparisans.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons