‘Error’ terms have a distinctive role in microeconometric models; because they represent unobservable individual preferences or, more generally, unobserved response patterns. Since the predictive power of a microeconometric model is derived from the observable data (the ‘signals’), it cannot be raised beyond the limits of the inherent individual deviations (the ‘noise’). Hence, the signal-to-noise relation between observable and unobservable factors is fundamental to microeconometrics, and also central to our brief study.
We address readers who are not necessarily specialists in econometrics but want to apply advanced methods. For that reason, the discussion of methods and examples starts from basic principles of statistics. Our strategy is to present the essence of a model class at hand, followed by remarks about pitfalls that may lead to misspecification, invalid tests, or unreliable forecasts. We add suggestions on how to improve the predictive power by choice of data and construction of adequate indices. The article starts with the linear regression model and some extensions, turning then to qualitative response and to selected topics of semiparametric survival analysis. The article concludes with examples thereby building a bridge to current developments in nonparametric and spatial econometrics.
‘Error’ terms have a distinctive role in microeconometric models; because they represent unobservable individual preferences or, more generally, unobserved response patterns. Since the predictive power of a microeconometric model is derived from the observable data (the ‘signals’), it cannot be raised beyond the limits of the inherent individual deviations (the ‘noise’). Hence, the signal-to-noise relation between observable and unobservable factors is fundamental to microeconometrics, and also central to our brief study.We address readers who are not necessarily specialists in econometrics but want to apply advanced methods. For that reason, the discussion of methods and examples starts from basic principles of statistics. Our strategy is to present the essence of a model class at hand, followed by remarks about pitfalls that may lead to misspecification, invalid tests, or unreliable forecasts. We add suggestions on how to improve the predictive power by choice of data and construction of adequate indices. The article starts with the linear regression model and some extensions, turning then to qualitative response and to selected topics of semiparametric survival analysis. The article concludes with examples thereby building a bridge to current developments in nonparametric and spatial econometrics.
การแปล กรุณารอสักครู่..