ScholarWorks@숙명여자대학교

초록

Symbolic data have become increasingly popular in the era of big data. In this paper, we consider density estimation and regression for interval-valued data, a special type of symbolic data, common in astronomy and official statistics. We propose kernel estimators with adaptive bandwidths to account for variability of each interval. Specifically, we derive cross-validation bandwidth selectors for density estimation and extend the Nadaraya-Watson estimator for regression with interval data. We assess the performance of the proposed methods in comparison with existing kernel methods by extensive simulation studies and real data analysis.

키워드