Comparison of estiamators in successive sampling models
Abstract
Successive sampling is a type of sampling which is carried out on two
or more occasions. Such sampling could either be on the same items or
different items on the same population which is sampled repeatedly and the
same study variables are measured at each occasion. Although much work
has been done on successive sampling, most of the work has however been
confined to using ratio and regression estimates on the matched portions of
the sample on the current occasion. Less work has been done on the
comparison of efficiencies of the estimators when sampling on more than two
occasion as well as developing second-order autoregressive process time
series estimators. The objectives of the study is to compute ratio and
regression estimates in successive sampling using forest data and compare
the efficiency of ratio and regression estimators. Estimators for successive
sampling based on a second order autoregressive time series model will also
be developed.
The results show that the ratio estimators over-estimate the population
mean, while the simple mean under-estimate it. The gain in precision in using
the ratio estimator over the simple mean is 9.57 %, and over the regression
estimators is 4.80%. Comparison of ratio and regression estimators showed
that the gain in precision using regression estimator over ratio estimator is
4.78 %. A survey carried out on a single occasion cannot give better
information of the natur-e of the changes or rate of such changes, because
population element may change over time. A major advantage of this study is
that since most populations are dynamic, regression and time series
estimators would be more appropriate so. as to avoid loss of useful
information.
Citation
M.Sc (Mathematical Statistics) ThesisSponsorhip
University of NairobiPublisher
School of Mathematics, University of Nairobi
Description
Master of Science Thesis