Geographically weighted ridge regression estimator of finite population mean to tackle multicollinearity in survey sampling Udgata Ashis Ranjan1,2, Rai Anil3, Sahoo Prachi Misra2, Ahmad Tauqueer2, Biswas Ankur2,* 1Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi-110012 2ICAR-Indian Agricultural Statistics Research Institute, New Delhi-110012 3Indian Council of Agricultural Research, New Delhi-110012 *Corresponding author email id: ankur.biswas@icar.gov.in
Online Published on 18 March, 2024. Abstract The utilization of auxiliary information in survey sampling significantly contributes to the estimation process. The presence of spatial information necessitates the exploration of a spatial model. The Geographically Weighted Regression (GWR), a spatial model regression has been widely applied to many practical fields for exploring spatial non stationarity. However, the occurrence of multicollinearity among the local variables in GWR model affect the estimation process. In this study, a new Geographically Weighted Ridge regression (GWRR) estimator has been proposed by taking care of the effect of multicollinearity in survey data. Proposed estimator performs better than other traditional estimators in terms of RRMSE value. Top Keywords Geographically weighted regression (GWR), Multicollinearity, Spatial model. Top |