Nonparametric Estimation of Regression Models with Unknown Error Distributions

The non-experimental nature of economic data makes it important that econometricians are able to estimate the relationship between variables through regression models. The major problem is how we can estimate such a relationship without specifying the error density. This project aims to simultaneously estimate such a relationship and the error density using the nonparametric regression model. The significance of this project is that it allows the data to choose bandwidth parameters that control the performance of the estimated regression function and error density. This project will enable econometricians and statisticians to investigate relationships between variables using more robust regression models without specifying error density.