Estimation of the average yield of cotton using outlier robust geographically weighted regression approach

Abstract

The General Crop Estimation Survey (GCES) scheme requires a large number of Crop Cutting Experiments (CCEs) to be conducted to get a reliable estimate below the district level. However, conducting a large number of CCEs imposes a financial burden on Govt. agencies. Additionally, largescale surveys like GCES often result in many outlier observations in the CCE data. To address this issue,this study was conducted to estimate the yield rate of cotton with a relatively fewer number of CCEs than the GCES scheme using the proposed Outlier Robust Geographically Weighted Regression (ORGWR) approach. Validation of the proposed methodology was done using the real CCE dataset of Amravati district for the 2012-13 agriculture year in Maharashtra. In this approach,the number of CCEs conducted for GCES scheme was reduced, and then this reduced number of the CCEs can be predicted using the proposed ORGWR approach. The predicted CCEs and the incomplete CCEs data are then combined to form a complete dataset. This complete dataset is used to calculate the crop yield accurately. The study conducted a comparison between the ORGWR approach and GCES methodology for estimating the average yield of cotton. The results showed that the ORGWR approach, when used with a lesser number of CCEs, yielded estimates that were almost equivalent to those obtained using the GCES methodology with the complete dataset. Moreover, the standard error of the estimate was reliable, indicating the validity of the results.