期刊名称：Journal of Statistical and Econometric Methods
印刷版ISSN：2241-0384
电子版ISSN：2241-0376
出版年度：2018
卷号：7
期号：3
语种：English
出版社：Scienpress Ltd
摘要：Paneldata models assume that parameters are common to each subject, that assumption isnot satisfied in many cases. The slope heterogeneity problem may be solved byobtaining groups where the slope parameters are heterogeneous across groups buthomogeneous within groups, followed by panel data theory within each group. Inthis paper, an algorithm to determine clusters of subjects is discussed; theclustering is achieved by checking whether confidence intervals from differentsubjects overlap or not. The number of groups is determined based on the datavariability. The clusters are useful by themselves to analyze the similarbehavior of subjects. Monte Carlo simulations were performed to examine theproperties of the methodology considered. Finally, clusters of countries withsimilar GDP per capita trend were obtained.
关键词：Panel data models; Clustering of subjects; Parameter homogeneity test; Confidenceintervals
