Sharma, M.K. (Professer)Berhanu, Damtew2018-06-252023-11-092018-06-252023-11-092016-06http://10.90.10.223:4000/handle/123456789/3195Small area estimation (SAE) plays an important role in survey sampling due to the growing demands of such statistics by both government and private sectors for various decision making and planning purposes. The main purpose of this study was to produce small area estimates of maize yield at wereda level based on a unit level mixed effect model approach and compare the results with direct estimator and other model based indirect estimators. In order to achieve this goal we used annual agricultural sample survey data from Central Statistical Agency (CSA) and other set of spatial auxiliary information from CSA, Ministry of Agriculture and Natural Resources (MoANR) and web sites. The AGSS data for 238 weredas of Oromia Region included in the sample survey was used to compute direct and model based estimators. The model based estimators compared were: EBLUP_B based on unit level mixed model, SEBLUP_A based on spatial Fay Herriot’s model, SYN_SLM based on simultaneous autoregressive lag dependent linear model and SYN_SACLM based on spatial simultaneous autoregressive SAC linear model. Using four diagnostic metrics the study revealed in general that EBLUP_B estimator show better performance than other estimators. Key words and phrases: Small area estimator, direct estimator, small area model, wereda, empirical best linear unbiased predictor, spatial dependence, maize yield, weight matrixenSmall area estimatordirect estimator, small area modelweredaempirical best linear unbiased predictorspatial dependencemaize yieldweight matrixThesis