ECON6300/7320/8300 Advanced Microeconometrics Non-Parametric methods
Christiern Rose 1University of Queensland
Practical 10 May 2019
1/4
Introduction
This class will review: Histograms
Kernel density
Non-parametric regression – Lowess/K-nearest neighbours
We begin with a demonstration from Chapter 9 of Microeconometrics: Methods and Applications.
We move on to a practical in which we estimate non-parametric Engel curves.
2/4
Practical (1)
We have the same World Bank data as last week
Thus far we have assumed log-linear Engel curves
Now we look more closely at the functional form
We focus on regressions of log-medical expenditure on log-total expenditure with no controls.
3/4
Practical (2)
1. Load, describe and summarise the data.
2. Estimate a histogram and kernel density of log-medical expenditure using Silberman’s optimal bandwidth and the Epanechnikov kernel.
3. Estimate a linear Engel curve for log-medical expenditure on log-total expenditure
4. Use a parametric approach to estimate a non-linear Engel curve.
5. Perform a hypothesis test of non-linearity in your parametric specification. What do you conclude?
6. Estimate a non-parametric Engel curve.
7. Plot the linear, parametric non-linear, and non-parametric
Engel curves on the same graph.
8. Based on your analysis so far, is linearity a reasonable assumption?
4/4