|
Conference on Quantitative
Social Science Research Using R
|
|
Keynote speaker: Prof. Roger Koenker, Dept of Economics, Univ. of Illinois
|
|
Quantile Regression: A Gentle Introduction to Estimating Models
for Conditional Quantiles in R
|
It is frequently desirable to model conditional quantile functions as a complement
to classical least-squares fitting of conditional mean models. Models with additive
nonparametric effects offer an crucial dimension reduction device for the non-parametric
component of such models. Shrinkage methods based on total variation roughness
penalties have proven to be particularly useful for quantile models of this type,
and are easily combined with more familiar "lasso" penalties on linear covariate effects.
An R implementation of of these methods will be described and illustrated with a model of
childhood malnutrition in India.
|
|
|
|
