Improved Estimators of Coefficient of Variation in a Finite Population

Coefficient of Variation (C.V) is a unitless measure of dispersion. Hence it is widely used in many scientific and social investigations. Although a lot of work has been done concerning C.V in the infinite population models, it has been neglected in the finite populations. Many areas of applications of C.V involves the finite populations like the use in official statistics and economic surveys of the World Bank. This has motivated us to propose six new estimators of the population C.V. In finite population studies regression estimators are widely used and the idea is exploited to propose the new estimators. Three of the proposed estimators are the regression estimators of the C.V for the study variable while the other three estimators makes use of the regression estimators of population mean and variance to estimate the ratio Y y , the population C.V for the study variable. The bias and mean square error (MSE) of these estimators were derived for the simple random sampling design. The performance of these estimators is compared using two real life data sets. The simulation is carried out to compare the estimators in terms of coverage probability and the length of the confidence interval. The small sample comparison indicates that two of the proposed estimators perform better than the sample C.V. The regression estimator using the information on the Population C.V of the auxiliary variable emerges as the best estimator.