MEASURING OVERFITTING IN NONLINEAR MODELS: A NEW METHOD AND AN APPLICATION TO HEALTH EXPENDITURES

摘要: When fitting an econometric model, it is well known that we pick up part of the idiosyncratic characteristics data along with systematic relationship between dependent and explanatory variables. This phenomenon as overfitting generally occurs when a model excessively complex relative to amount available. Overfitting major threat regression analysis in terms both inference prediction. We start by showing Copas measure becomes confounded shrinkage or expansion arising from in-sample bias applied untransformed scale nonlinear models, which typically interest assessing behaviors analyzing policies. then propose new expressed on immune this problem. also show how respective contributions overall predictive applying estimated data. finally illustrate properties our through simulation study real-data illustration based inpatient healthcare expenditure data, shows distinctions can be important.