<![CDATA[This study is motivated by the non-response problem in the Crop Cutting Survey conducted by the BPS-Statistics Indonesia as the official statistics provider. BPS has a vision of providing quality statistical data for advanced Indonesia. Handling non-response is essential to supporting this vision because non-response can potentially cause some sample characteristics to be unrepresented. This study proposed a non-response data imputation technique through statistical modeling. The proposed model was an additive model with the addition of geospatial smoothing functions of thin plate regression splines (TP) and Gaussian process (GP). Selection of the best model based on the smallest MSEP of 1000 iterations. Then we compared the average rice productivity between listwise deletion and imputation techniques through three scenarios of non-response data. The results showed that the model with the addition of the GP smoothing function gave the best performance with the smallest MSEP. The other results showed that the imputation method of non-response data is better than ignoring non-response. BPS can consider the imputation method to improve the quality of official statistics on rice productivity.]]>
Copyrights © 2023
