johnjosephhorton

# Find all your R files & get just the package names, uniquely 
find ../analysis/ -name *R -print0 | xargs -0 grep -h 'library*' | sed 's/library//g' | sed 's/#//g' | sed 's/ //g' | sort | uniq|\
 sed 's/(//g' | sed 's/)//g' > package_list.txt
 
# Install each package 
cat package_list.txt | xargs -I {} Rscript -e 'install.packages("{}")'

johnjosephhorton

#!/usr/bin/Rscript --vanilla

suppressPackageStartupMessages({
library(magrittr)
library(dplyr)
library(foreign)
library(plm)
library(reshape2)
library(stargazer)
library(ggplot2)

johnjosephhorton

project_name = hot_towel

ts := $(shell /bin/date "+%Y-%m-%d---%H-%M-%S")

params =
# DONE 
params += params_exp_details.tex
params_exp_details.tex: ../analysis/params_exp_details.R
	cd ../analysis && ./params_exp_details.R

johnjosephhorton

library(XML)
library(ggplot2)

df <- readHTMLTable("http://projects.dailycal.org/paychecker/departments/")[[1]]

DeMoney <- function(x) as.numeric(gsub(",", "", gsub("\\$", "", as.character(x))))

money.columns <- c("All", "Professor", "Associate professor", "Assistant professor", 
                   "Lecturer")

johnjosephhorton

library(geofacet) # my forked version 
library(ggplot2)
library(jsonlite)
library(ggrepel)

# grab top cities 
cities <- fromJSON("https://gist.githubusercontent.com/Miserlou/c5cd8364bf9b2420bb29/raw/2bf258763cdddd704f8ffd3ea9a3e81d25e2c6f6/cities.json")

n <- 50
x <- cities$longitude[1:n]

johnjosephhorton

library(lfe)
library(stargazer)

set.seed(6152011)

# Number of cities                                                                                                       
L <- 1000

# Number of industries                                                                                                        
K <- 3

johnjosephhorton

SimExperiment <- function(num.groups, sample.size, beta, randomize.by.group = FALSE){
    "Simulate running an analysis of the experiment with linear regression"
    df <- CreateClusteredData(num.groups, sample.size, beta, randomize.by.group = randomize.by.group)
    m <- lm(y ~ trt, data = df)
    c(as.numeric(coef(m)[2]), as.numeric(sqrt(diag(vcov(m))[2])))
}

johnjosephhorton

CreateClusteredData <- function(num.groups, sample.size, beta, randomize.by.group = FALSE){
    individual <- 1:sample.size # index for individuals
    group <- sample(1:num.groups, size = sample.size, replace = TRUE) # randomly assign everyone to a group
    group.effect <- rnorm(num.groups) # create a group-specific effect
    eta <- group.effect[group] # give each individual their group-specific effect
    epsilon <- rnorm(sample.size) # given each individual an individual-specific effect
    if (randomize.by.group){
        group.trt.assignment <- rbinom(num.groups, 1, 0.5) 
        trt <- group.trt.assignment[group]
    } else {

johnjosephhorton

ts := $(shell /bin/date "+%Y-%m-%d---%H-%M-%S")

project.pdf: project.bib project.tex
        pdflatex -interaction=nonstopmode project
        bibtex project
        pdflatex -interaction=nonstopmode project
        pdflatex -interaction=nonstopmode project

backup:
        tar --exclude='../backups' -zcvf ../backups/project_$(ts).tar.gz ../

johnjosephhorton

library(ggplot2)

# simulate experiments using stratification and randomization and report
# absolute difference in estimates from true treatment effect, tau. 

simExp <- function(n, tau, beta = 3){
    x <- sort(runif(n))
    W.rand <- rbinom(n,1,0.5)
    W.strat <- rep(c(0,1), n/2)
    epsilon <- rnorm(n, 0, 0.5)