In potterzot/rkclass: K-class regressions (Two-Stage Least Squares, LIML, Fuller)

knitr::opts_chunk$set(eval = TRUE, include = TRUE, echo = TRUE)

Overview

This is a simple example of how to use the rkclass package to get you up and running quickly. We will do a simplified regression of the type found in Angrist and Krueger [-@AngristKrueger1991].

rkclass comes with a dataset from that paper that can be loaded with:

#load the package
library(rkclass)

qob = data(qob)

A simple variant of a regression from Angrist and Krueger is:

#formula = LWKLYWGE ~ EDUC + AGE + AGESQ | . - EDUC + as.factor(YOB)
#m.kclass = kclass(formula, data=qob)
#summary(m.kclass)

This output should match (and is tested against it on build) output from AER::ivreg

#m.aer = AER::ivreg(formula, data=qob)
#summary(m.aer)

Additional Options

rkclass makes several things easy that are more difficult or not supported in AER. Being a general k-class estimator, you can run a number of models other than the default LIML model. These are specified by the parameter model.type, for which you can specify "LIML", "OLS", "FULLER", "TSLS", or "KCLASS". "FULLER" requires an additional alpha parameter.

#m.liml = kclass(formula, data=qob, model.type="LIML")
#summary(m.liml)
#m.fuller = kclass(formula, data=qob, model.type="FULLER", alpha=1)
#summar(m.fuller)

The k parameter can also be specified directly, though that's not generally recommended, and is overridden if model.type is set to anything other than "KCLASS".

potterzot/rkclass documentation built on May 25, 2019, 11:24 a.m.