Research Blog

I’m working on a research replication projects for my Stats class and have run into a little trouble converting a Stata DO-File to [R] script.  I’m working to extend the original research of Craig Depken and Trisha Bezmen which was presented in “Influences on software piracy: Evidence from the various United States“.  Thankfully Dr. Depken was able to provide me with the original data set, Stata DO-File, and the variable descriptions (Shared below with permission from Dr. Depken, be sure to thank him if you use this data).

Now for the problems I’m facing.  The original DO-File uses these commands:

sum

reg piracy lninc sfindex salestax unemp yr00 yr01,r

reg piracy lninc unemp statetax sfindex yr00 yr01,r

xtreg piracy lninc unemp statetax sfindex yr00 yr01,fe

xtreg piracy lninc unemp salestax sfindex yr00 yr01,fe

xtreg piracy lninc unemp statetax sfindex yr00 yr01,re

est store random1

xttest0

quiet: xtreg piracy lninc unemp statetax sfindex yr00 yr01,fe

est store fixed1

hausman  fixed1 random1

xtreg piracy lninc unemp salestax sfindex yr00 yr01,re

est store random2

xttest0

quiet: xtreg piracy lninc unemp salestax sfindex yr00 yr01,fe

est store fixed2

hausman fixed2 random2

At this point I have created this [R] script based on the commands from the original DO-File:

### Created: March 1, 2011

### Author: Frank Jones

### Last Saved: March 18, 2011

### Piracy Replication

### Generaly good practice to begin R script files with these commands

# clear any objects in memory

rm(list=ls(all=TRUE))

# file location

file.location <- c(“/home/frank/Dropbox/UNC/Spring2011/POLI 784.974/Research Replication/My Work/piracydata-jones.dta”)

# load data

require (foreign)

originalData <- read.dta (file.location, convert.dates=TRUE, convert.factors=TRUE, missing.type=TRUE, convert.underscore=FALSE)

# load packages

install.packages(c(“nlme”,”MASS”))

# run once

library(nlme) # used for computing fixed- and random-effect models (lme4 is an alternative package)

library(MASS) # used for computing robust regression models

# sum

# reg piracy lninc sfindex salestax unemp yr00 yr01,r

reg1 <- rlm(originalData$piracy ~ originalData$lninc + originalData$sfindex + originalData$salestax + originalData$unemp + originalData$yr00 + originalData$yr01)

# reg piracy lninc unemp statetax sfindex yr00 yr01,r

reg2 <- rlm(originalData$piracy ~ originalData$lninc + originalData$unemp + originalData$statetax + originalData$sfindex + originalData$yr00 + originalData$yr01)

# xtreg piracy lninc unemp statetax sfindex yr00 yr01,fe

originalData.Grouped <- groupedData(piracy ~ lninc + unemp + statetax + sfindex + yr00 + yr01 | state, data=originalData) #lme function requires grouped data (grouped here by state)

reg3 <- lme(fixed = piracy ~ lninc + unemp + statetax + sfindex + yr00 + yr01, data = originalData.Grouped)

# xtreg piracy lninc unemp salestax sfindex yr00 yr01,fe

reg4 <- lme(fixed = piracy ~ lninc + unemp + salestax + sfindex + yr00 + yr01, data = originalData.Grouped)

# xtreg piracy lninc unemp statetax sfindex yr00 yr01,re

reg5 <- lme(random = piracy ~ lninc + unemp + statetax + sfindex + yr00 + yr01, data = originalData.Grouped)

# est store random1

# xttest0

# quiet: xtreg piracy lninc unemp statetax sfindex yr00 yr01,fe

# est store fixed1

# hausman  fixed1 random1

# xtreg piracy lninc unemp salestax sfindex yr00 yr01,re

# est store random2

# xttest0

# quiet: xtreg piracy lninc unemp salestax sfindex yr00 yr01,fe

# est store fixed2

# hausman fixed2 random2

I am having problems with finding an [R] command which calculates fixed- and random-effect models similar to the xtreg function found in Stata.  I’ll post my updated version of this [R] conversion along with the results of my replication once it’s done.  Any suggestions for [R] functions that are similar to xtreg are greatly appreciated as well as any other comments you may have about this project.

The Data provided to my by Dr Depken:

CSV Data

DTA Data

DO-File

Variable Descriptions

My current [R] script:

3/18/11 [R] script