Estimação de regressões aditivas via backfitting e integração marginal: performance em pequenas amostras

• Autor(a) principal: Silva, Fernando Augusto Boeira Sabino da
• Data de Publicação: 2001
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: https://hdl.handle.net/10438/179
• topic: Economia; Análise de regressão; Monte Carlo, Método de

In this thesis we conduct a Monte Carlo investigation to reveal some characteristics of the small sample distributions of the Backfitting (B) and Marginal Integration (MI) estimators for an additive bivariate regression. We are particularly interested in providing some evidence on how different data driven window width estimation procedures, such as some plug in methods impact the small sample properties of the MI and B estimators. We are also interested in providing evidence on the behavior of how the differente window widths estimators impact the optimal sequence of win- dow widths that minimizes a chosen loss function. The impact of ignor- ing regressor dependency on window width estimation is also investigated. This is common practice and should impact estimators' performance. Be- sides, nowadays there no available statistical/econometrical packages that perform estimation of additive regression by Backfitting and Marginal In- tegration. It's an objective of our dissertation the creation of routines in Gauss for the practical implementation of these estimators. Ultimately, differently from what occurs at the present time, when the utilization of the B e MI estimators is done in a way completely ad-hoc, our objective is to provide applied researches with information that allows for a more accurate comparison of these two competing alternatives in a finite sample setting.