In recent years, the integration of data from multiple surveys has gained popularity in sample survey research. Broad-scale survey programmes often face practical challenges, including high survey data collection costs and increased demand for disaggregated statistics. These challenges were addressed through the integration of multiple survey data sets. Spatial parameters play a significant role in many surveys, leading to spatial non-stationarity, a phenomenon where the relationship between the study variable and covariates varies across locations. To address this issue and obtain more detailed information, sampling methodologies capable of efficiently handling spatial non-stationarity and integrating spatially referenced data are needed. This research introduced a Geographically Weighted Regression model-based Spatially Integrated estimator designed to estimate finite population totals by combining data from two separate surveys. The statistical characteristics of the proposed estimator were evaluated through an empirical design-based simulation study using real data from Crop Cutting Experiments and model-based spatial simulation studies, which demonstrated its superiority over traditional design-based estimators. Additionally, a spatial bootstrap method was developed to estimate the variance of the integrated estimator. AMS Subject Classification (2000): 62 D05
Nobin Chandra Paul (Tue,) studied this question.