Some Issues on Modeling Unobservable Variables, III

点击翻译按钮获取中文摘要

Statistics and econometrics are mainly concerned with the quantitative measurement and analysis of real world phenomena. Typically, it is assumed that the data for the variables of interest are available and are measured accurately. In reality, errors of measurement are common because of reporting errors or because behavioral responses are in part determined by unobserved characteristics of agents; thus the variables of interest are latent or unobservable. This project is concerned with developing an analytical framework to extract information from data that is subject to either random or systematic measurement errors. The objective is to refine the methodology to establish empirically valid inferences from imperfectly measured data. In particular, we shall study issues of (i) random measurement errors in time series data that contain (stochastic) trends; (ii) systematic errors due to biased or exaggerated responses of sample respondents or grouping of individual observations into categories for the sake of confidentiality or convenience of collecting data; (iii) extracting information on the implicit market for characteristics (or attributes) from the "hedonic" approach developed by economists to simplify the analysis of differentiated products and quality changes over time; and (iv) modeling unobserved individual characteristics in panel data analysis which directs its focus on individual outcomes.