In nateybear/longhorn-rstata: A Package for UT Students who Hate Stata

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

An R Package for Stubborn Longhorn Economists

Authors: Nate Hattersley and Chase Wiedemann

This is a package that attempts to mimic the full suite of Stata used in the ECO394M Econometrics class. The goal is to make the usage of this R package as close to Stata as possible, and to provide the tools so that the naive user can replicate Stata in a non-licensed language. It is not necessarily the quickest way to do something in R. For that, dig into the code here on GitHub.

Installation

This code is hosted on GitHub. Use the devtools package to download the package:

# install.packages("devtools")
devtools::install_github("nateybear/longhorn-rstata")

Example

Let's walk through an example using the HTV.DTA dataset from class. I will briefly run through all of the functions that are in the package. If you need more documentation on, for example, the gen function, then type ?gen in your R console.

use

The use command will load a dataset. Like Stata, rstata is meant to work with one dataset at a time. Let's browse our computer for the HTV.DTA file and load it:

use()

library(rstata)
## basic example code
trash <- capture.output(use(htv))
suppressMessages(library(pryr))

Now we can use the rest of the rstata functions.Note that the dataset is attached to your R environment, so you can use the columns of HTV.DTA directly.

summary(abil)

Note you can also browse through your Environment tab in RStudio and view the values in the dataset:

regr

The regr command fits a model just like Stata. Note the tilde instead of the equal sign from Stata. Separate options with multiple commas.

This will get filled out with more options as we go along. Have patience :)

regr(educ ~ motheduc + fatheduc + abil)

b_

After the regression is run, you can access the fitted coefficients with b_, just like Stata's _b. Note that this is a function in R, so use parentheses:

b_(motheduc)

gen

Use the gen function to create new variables and attach to your dataset. As an example, let's create a more complex model that has an interaction effect. Then we're going to compute the partial effect of mother's education on the dataset and plot a histogram.

regr(educ ~ motheduc*fatheduc + abil)

This includes our first "weird" variable name. Make sure to surround the variable for the interaction term with a backtick (`). Let's compute the partial effect now and plot a histogram:

gen(pe_fatheduc = b_(motheduc) + fatheduc * b_(`motheduc:fatheduc`))
hist(pe_fatheduc, breaks = 20)

testnl

Test multiple (possibly non-linear) restrictions jointly. This uses the delta method to get the variance of non-linear functions of the beta parameters, and then computes a Wald chi-square test statistic.

Note that this gives an exact answer for linear restrictions thanks to symbolic differentiation. Hence there is no loss of precision if you are testing something like a single linear restriction:

# test that b_(motheduc) = 0. Compare the pval to the output of regr.
# TODO this uses robust standard error, need to integrate with regr!!!
testnl(motheduc)

You can do a partial F-test:

testnl(`motheduc:fatheduc`, abil)

Or you can test a non-linear restriction:

testnl(sqrt(motheduc^2 + fatheduc^2) - 1)

nateybear/longhorn-rstata documentation built on Oct. 21, 2020, 2:14 a.m.