Abstract:
|
The auxiliary information that is often found to be useful for small area modeling may be contained in different datasets. This situation may arise due to ownership and data collection schemes, privacy and confidentiality considerations, or data security requirements. Also, in many modern instances, the data used for small area modeling is high-dimensional in nature. Additionally, such big data may contain outliers and aberrant observations, consequently most statistical techniques need to be used with caution when analyzing such big data. In this work, we present a robust and computationally simple technique for linking high-dimensional datasets. We present theoretical foundations for our proposed methodology, and illustrate using numeric examples.
|