Event Date
Location
IFPRI - Dakar
Overview
This course will expose students to the current practice of applied nonparametric methods. These methods are designed to overcome misspecification issues that dominate applied economic research. Using nonparametric methods can shed light on new issues that were previously hidden due to ad hoc model specifications imposed in the analysis. This class will provide hands on instruction to students in state-of-the-art nonparametric methods designed specifically to assist students with empirical issues that arise in the application of these techniques. The basic concepts of smoothing will be covered, complemented with numerous examples and illustrations through the open source software R. This combination of class instruction and computer tutorials will assist students in developing a sound foundation to apply nonparametric methods in their own research agendas.
Applications must be submitted by June 10, 2015
Objectives
Participants of this course should leave with the ability to understand the nuances of nonparametric estimation and inference and the empirical implications that manifest. Further, participants should be able to successfully integrate their data into R and construct nonparametric estimates for their models, conduct inference and rigorously interpret these results to provide sound policy insights. All methods discussed will be accompanied with corresponding R code, data and documentation to the literature at large making it easy for participants to follow along in the class as well as a check once the class has ended and they are engaged in their own analysis.
Course Outline
Day 1:
- Introduction to R using the basic linear regression model (all notes, data and examples will be provided)
- Motivation for using nonparametric methods
- Kernel smoothing and density estimation
- Selection of the smoothing parameter
- High dimensional settings and the curse of dimensionality
- inference for estimated densities
- Computer Tutorial
Day 2:
- Kernel regression: Motivation and intuition
- Kernel smoothing and OLS
- Bandwidth Selection
- Irrelevant variables in the model: the case of LSCV
- Hypothesis testing in regression settings
- Hypothesis testing continued
- Presenting results from nonparametric regression
- Computer Tutorial
Day 3:
- Semiparametric Methods: Motivation and intuition
- The partly linear model
- The single index model
- The additively separable model
- The smooth coefficient model
- Computer Tutorial
Day 4:
- Incorporating discrete variables: motivation and intuition
- The discrete kernel density estimator
- Irrelevant discrete variables
- Accounting for panel data
- A random e ffects framework estimator
- A fixed eff ects framework estimator
- A fully nonparametric dynamic panel data estimator
- Computer Tutorial
Day 5:
- Endogeneity in nonparametric models
- A nonparametric IV estimator
- A semiparametric IV estimator
- What are weak instruments in a nonparametric setting?
- Imposing smoothness constraints in nonparametric setting
- Inference in constrained nonparametric settings
- A nonparametric stochastic frontier estimator
- Computer Tutorial
Pre-Requisites
The course level is appropriate for participants with a background in economics, statistics, mathematics, and/or public policy. A strong background in quantitative analysis is required. Basic knowledge of the statistical software R is desirable. A general fluency in the statistical/econometric lingo at the (post-) doctoral level (hopefully in a non-statistics/econometric discipline) is required. More speci fically, the Law of Large Numbers and the Central Limit Theorem should be understood.
Software Requirements
This course will heavily leverage implementation in R, a powerful statistical software package that is freely available. R possesses the facilities to implement an impressive array of nonparametric estimators and tests as well as to serve as an interface for data manipulation, making it an ideal choice when discussing the application of nonparametric methods. Moreover, R's real strength is that users can readily and easily construct their own estimators and tests so that canned approaches do not need to be relied on, allowing users to stand on their legs when conducting empirical research.
Applications
In order to apply for this course, AGRODEP members must complete the following by June 10, 2015:
If you would like to practice using Stata before taking the proficiency test, please review the modules below. Information included covers Stata use for beginners, linear regressions, bivariate regressions, and panel data. You will need to know this information to successfully complete the test.
- Training Module 1: Introduction to Stata
- Training Module 2: Basic Data Management, Graphs, and Log-Files
- Training Module 3: Linear Regressions
- Training Module 4: Bivariate Regressions
- Training Module 5: Panel Data Regressions
Real Data Nonparametric Applications
Every participant is allowed to submit one application, no later than 4 weeks before the course. A selection will be made from the submitted applications to discuss in detail in the class and to illustrate the practical pitfalls that are encountered with real data. Additionally, applications and data sets taken from published research will also be made available for the class to provides participants with a truly hands on approach to nonparametric econometrics.
Instructor
Christopher F. Parmeter is an Associate Professor in the Economics Department at the School of Business Administration at the University of Miami. His area of expertise is in applied econometrics with special interests in semi- and nonparametric methods, benefit transfers, meta-analysis and efficiency analysis. Dr. Parmeter has coauthored 30 peer reviewed scientific articles in leading econometric and applied economics journals.