Anil RaiYOGITA GHARDE2016-12-012016-12-012011http://krishikosh.egranth.ac.in/handle/1/88550T-8512If parameter of interest is geographical in nature, it is always desirable to incorporate the spatial structure into the model for estimation of parameter of interest. Therefore, incorporating the spatial effects in the model through both simultaneously autoregressive error process and geographically weighted regression approach, it is possible to improve the precision of small area estimates. Further, it is expected to have more efficient small area estimates using prior knowledge about the parameter of interest i.e. using Bayesian approach in spatial small area models. In this study, an approach has been developed to incorporate the spatial information in the random area effect present in the unit level small area model. This approach is based on geographically weighted regression technique which incorporates the spatial information in the model. Small area estimators were obtained along with MSE using this model. The comparisons of proposed estimators with non-spatial estimators have been made using simulated population for 15 different small areas. Three approaches were used to assign spatial weights i.e. Neighbourhood Criteria method, Gaussian-Decay method and Spherical variogram approach. Again an attempt has been made to obtain small area estimates for above developed spatial model under hierarchical Bayes (HB) framework. The spatial model proposed by Chandra et al. (2007) has also been studied in Hierarchical Bayes framework. Under these, the small area mean is estimated by its posterior mean and its precision is measured by its posterior variance. Sensitivity analysis has also been done for both the model using three approaches of incorporating spatial effects.Thesis