Introduction

This page describes the general RA guidelines we employ for our projects. The purpose of the guidelines is to make it easier to focus on doing replicable and properly documented work. The guidelines draw heavily on

Code and Data for the Social Sciences: A Practitioners Guide. Gentzkow and Shapiro (2014). Link
- The document contains a general introduction to reproducibility challenges that many social science researchers have faced and how the two authors have tried to solve them in their lab. Emphasizes coding conventions, directories, and version control.
Lab Manual. Gentzkow and Shapiro. Link
- This supplements their (2014) document. It includes information on workflows, coding, data handling, and paper and slide production. We generally follow their workflow approaches (e.g., using GitHub issues to assign and resolve tasks).

How to get started: (Talk with supervisor about which tasks makes sense for you)

Read Gentzkow and Shapiro (2014)
Read this document.
- You do not have to follow links the first time you read the document. They are there for reference.
Install the necessary software on your local computer as described under Software/Required.
Set up git on your computer and a GitHub account as described under Version Control.
Ask your supervisor to add you to the relevant projects on GitHub and Zotero.
Get to know R
- Getting started: Read Hans Henrik Sievertsen's Introduction to R and solve the associated exercises.
- data.table: Read through Atrebas' introduction to using the data.table package and try out the commands in your own R script.
  - The document describes basics like viewing data, subsetting, creating new variables, and using the .SD capability.
- exercises: Go through the r_introduction.qmd quarto document that you can find on this Github page.
- duckdb for large parquet files: Go through Grant McDermott's introduction to using duckdb, a lightweight database system, to work with large parquet files without having to load them into r.
  - duckdb + dplyr verbs is often substantially faster than loading data into R and using dplyr or data.table for (1) doing simple data wrangling, and (2) doing straightforward descriptive statistics such as counts, summary statistics, etc.

After you have finished these steps, we will assign tasks to you via the Github project(s).

Workflow

Tasks

We follow Gentzkow and Shapiro's workflow approach.
Tasks will be specified on GitHub under the relevant project as issues. We do this to keep track of open tasks, task notes and questions, and outputs. You can familiarize yourself with GitHub issues here.
- A supervisor will add you to the relevant project(s) on GitHub.
Each task will contain
- A description.
- A set of outcomes.
- A task supervisor.
- One or more task assignees.
When working on a task
- Keep documentation of your work. We suggest having a running_notes_(your initials).md document where you store your thoughts. A good practice is to add headlines with dates so that it becomes easier to go back to find thoughts related to a task you previously worked on.
Asking questions
- We encourage you to work independently but ask questions when you realize you are stuck or something seems unclear. We all get stuck. And it happens particularly often when we start working with the administrative data, so do come around and ask. Our experience is that RAs who ask questions early on are more productive in the long run.
- If you ask about clarifications or questions related to a task in person, add a note about the questions and answers to the GitHub issue so we can track the progress.
A task is closed by the task supervisor when
- The relevant outcomes have been created/reached.
- The task assignee has written a reply to the GitHub issue on how they completed the task and where relevant outcome files are located (e.g., the code file cleaning a bit of data or the note summarizing results).
- The task supervisor agrees that the task is completed.

Reporting and notes

We write notes and documents in markdown format with files ending in .md unless another format is required.
- Markdown documents can easily be compiled into Word, PDF (via LaTeX), HTML, beamer PDF slides, or other formats using pandoc or quarto.
- Searching (and replacing) across multiple markdown files for content using typical text editors is easy.
- Markdown files can be edited using most text editors. We suggest VSCode and Obsidian.
- Introduction to basic markdown syntax written by the developers behind the original markdown language (it comes in many flavors).
We keep personal running notes documents.
- These typically contain thoughts and drafts for notes and tasks.
- The document will typically be named running_notes_ and end in our initials, e.g., running_notes_je.md.

References

We use Zotero to maintain shared libraries with project references.
- A supervisor will add you to the relevant project(s) on Zotero.
We use the betterbiblatex extension for Zotero to export .bib files to projects or to Overleaf.
When adding a new reference to a Zotero project collection, pin the bibtex key.
- Right-click the reference and select better bibtex -> pin citekey.
- Remember to set up the better bibtex extension to use the correct citekey structure. See Software / Required / Zotero.

Writing papers

We use Overleaf to write papers in LaTeX format unless otherwise specified. Overleaf allows us to
- Work on the paper at the same time
- Integrate references from shared Zotero libraries

Version Control

We use the Git version control system to track changes to our notes and code. Git allows you to add, delete, or modify files, mark them as changed (committing), and finally add them to the project (pushing to the remote repository). All changes to files will then appear in the project Git history.
We use GitHub to host our remote repositories for non-sensitive project files.
- You can keep the local non-sensitive project files where you prefer.
- You will pull and push changes to this remote directory to update general project files.
- A supervisor will add you to the relevant project(s) on GitHub.
We use a local Git repository for content stored on secure servers.
A basic workflow for using Git is to Pull the latest version of project files from the remote repository.
1. Pull latest updates from the remote repository.
2. Make changes to the files you are working on.
3. Commit your changes.
  1. Add a reference to the task you are working on in the commit message if relevant. You do this by adding adding a reference to the task-number, e.g., #123 Add peer-group-control robustness regression - (text), in front of the commit message. (Read Chris Beams' article on good commit messages).
4. Pull the latest updates from the remote repository and resolve any conflicts.
5. Push your changes to the remote repository.
Guides
- Basic (takes about 20 minutes total and well worth it): Git Handbook, Understanding the GitHub Flow, Mastering Issues, Mastering Markdown
- Detailed: Pro Git, Version Control with Git, Chapters 4-9
- Writing good Git commit messages by Chris Beams.
Setup: (based on https://github.com/gslab-econ/lab-manual/wiki/Setup)
- Create a GitHub account and install the Git desktop/command line clients.

Project structure

Project storage

We generally store non-sensitive project files in private GitHub repositories (one per project). Using Github repositories ensures all project participants can access the relevant files.
- It also makes it easy to track file changes and revert to previous versions if necessary.
- We use the git version control system to pull and push files to the Github repositories.
- Supervisors will add you to the relevant project(s) on GitHub.
We store sensitive files on a secure UCPH server.
- Supervisors can give you access to relevant folders.
GDPR-sensitive microdata is typically stored on a secure server hosted at Statistics Denmark. This includes administrative data from Statistics Denmark and the Ministry of Education we use in our projects.
- Documentation on Statistics Denmark's researcher data access and storage is available here.
- If you work with sensitive microdata, your project supervisor will help you set up access.
- Before getting access, you are required to read and sign the internal UCPH ECON and Statistics Denmark guidelines on working with sensitive data.
- We follow the internal UCPH ECON guidelines on working with sensitive data, including what information can be downloaded from the secure server.

Directories

Our projects are typically distributed locally and on a secure server. The local project might contain notes, literature reviews, code that doesn't need to run on a secure server, and paper drafts. The secure server project location will exist when the project requires restricted access data (e.g., from Statistics Denmark).
We generally apply the rules from Gentzkow and Shapiro (2014), chapter 4:
1. Separate directories (folders) by function.
2. Separate files into inputs and outputs (and temporary files)
3. Make directories (folders) portable.
Each (sub)component of a project should have its separate folder. A project with a literature review, presentation files, and a paper (draft) should contain at least those folders.

lit_review/ 
presentations/
paper/

All components containing code should have at least a code, temp, and out folder. For example, assume that the simple project contains a simulation exercise written in R showing the consistency of an econometric estimator. The code file outputs the graph simulate_estimator_consistency_distribution.PDF. The folders could look like

consistency_simulation/
	code/ 
		simulate_estimator_consistency.R 
	temp/
	out/
		simulate_estimator_consistency_distribution.PDF
lit_review/ 
presentations/
paper/

We store raw data in a separate folder.
- Suppose we have more than one raw data set, for example, from Statistics Denmark and the Ministry of Education. We then separate them into subfolders with meaningful names and possibly a date of compilation so we can keep track of versions.

buildraw/ 
	dst/
		dat1.sas7bdat
		dat2.sas7bdat
consistency_simulation/
	code/ 
		simulate_estimator_consistency.R 
	temp/
	out/
		simulate_estimator_consistency_distribution.PDF
lit_review/ 
presentations/
paper/

Data formats

When possible, store data in .parquet format.
- .parquet files are typically substantially smaller than .csv or .dta files, so we can reduce our server footprint.
- Reading/writing .parquet files in R is typically a lot faster than .csv and .dta.
We can import and export .parquet files using R and python.
- Guide to using parquet with R by Apache.
- Guide to using parquet with R by Hadley Wickham.
In R, we will often use the rio package to import/export a dataset:

library(rio) 
dat = import("builddata/out/clean_bef.parquet")
dat |> export("builddata/out/clean_bef.parquet")

R projects

We use R projects to help RStudio determine where to run our R scripts.
- The R-project file must be located at the root of the project directory.
- Read about working with scripts and R projects here.

Code Conventions

General

Name scripts by what they do.
- Example: We have written code that cleans the raw BEF register data for use in subsequent analyses. The file (sh)could be named clean_bef.R.
Script outputs must contain the name of the producing script and be informative about the content.
- Example: Assume the file describe_main_sample.R outputs two summary tables in LaTeX format. One is a balance table with means and differences in means between treatment and control groups, and one contains general summary statistics for the full sample. These (sh)could be named describe_main_sample_balance.tex and describe_main_sample_summary.tex.
No line of code should be more than 100 characters long. All languages we work in allow you to break a logical line across multiple lines on the page (e.g., using /// in Stata or ... in Matlab). You may want your editor to show a “margin” of 100 characters.
Functions should not typically be longer than 200 lines.

R

We follow Google's R Style Guide
Exceptions to style guide:
- Separate using underscores, and keep lowercase. Example: .CalcMeans() should be calc_mean().
Use the rio package for data import/export.
- It supports many file types and generally uses the most efficient IO tool for importing/exporting the file format.
Use the data.table package for data wrangling when possible.
- Introduction to data.table from the authors of the package.
- Introduction for Stata users
- data.table chapter in Introduction to Data Science.
- data.table cheatsheet.
Use the fixest package by Laurent Berge for estimating most types of statistical models, particularly linear and IV models with fixed effects, when possible.
- It provides estimation tools typically much faster than other options in R and Stata.
- Linear, fixed effects, and 2SLS models can be estimated via feols().
- Documentation.
We typically use the modelsummary package for summarizing regression results and creating summary statistics tables outputted to latex or markdown format.
- An introduction is available here.
- When adding footnotes to modelsummary() tables, use the footnote() function from the kableExtra package with the options escape = F and threeparttable = T.
We use the here package when specifying paths to ensure all paths are read relative to the project folder.

Larger than memory datasets

We sometimes work with datasets that are larger-than-memory, meaning that loading them requires (nearly) all or more than the RAM available on servers. We can avoid loading data into R using the arrow and duckdb packages.
- Introduction to using arrow (with duckdb) in R.
- Introduction to using duckdb with dplyr functions.
Example of interacting with a parquet file using using the duckdb package dplyr verbs:

# load packages 
library(duckdk)
library(tidyverse)
library(arrow) 

# create and/or connect to existing database file (the name doesn't matter) 
db = dbConnect(duckdb(), "database.duckdb")

# open a connection to the parquet file and treat it as a duckdb database
#  select only the three of the variables in the dataset 
dat_db = open_dataset("path_to_data.parquet") %>% 
    to_duckdb() %>% 
    select(id, group = Group, salary)

# find number of groups 
dat_db %>% count(group) %>% count(n)

# Find average salary by group 
dat_db %>% summarise(ave_salary = mean(salary, na.rm = T), .by = c(group))

# Plot average salary by group number 
dat_db %>% 
    summarise(ave_salary = mean(salary, na.rm = T), .by = c(group)) %>% 
    ggplot(aes(group, ave_salary)) + 
    geom_point() 

# Load observations with salaries higher than 1,000,000 DKK into R 
dat_highsal = dat_db %>% 
    filter(salary > 1000000) %>% 
    collect() 

# Make R explain the database query it sends to duckdb when summarising salaries
dat_db %>% 
    summarise(ave_salary = mean(salary, na.rm = T), .by = c(group)) %>% 
    explain()

Guides to get started with R

How to get started:
- Read parts Hans Henrik Sievertsen's Introduction to R and solve the associated exercises.
- Read through Atrebas' introduction to using the data.table package.
  - The document describes basics like viewing data, subsetting, creating new variables, and using the .SD capability.
- Go through the r_introduction.qmd quarto document that you can find on this Github page.
Other great guides
- R for Data Science (2e) is an online and free goto reference for getting started with simple and more advanced R, including data IO with arrow, writing functions, and using quarto to communicate results.
  - Each chapter comes with great exercises.
  - The main author, Hadley Wickham, has been a driving force in developing R packages, including a majority of the tidyverse package since the 2000s.
- Hans Henrik Sievertsen's Applied Econometrics with R introduces using R for applied econometrics, including data cleaning, visualization, descriptive statistics, and regression analysis (using feols and modelsummary). This guide is great for when you've gotten the hang of basic R.
- Hans Henrik Sievertsen's Interactive introduction to R. It introduces basic data handling (loading data, modifying and merging datasets) and plotting.
ChatGPT often gives great solutions to coding problems!

Automation

We automate everything that can be automated. This implies writing scripts to do all data cleaning, analysis, and table formatting, and using build tools to run these scripts in the correct order.
We use the build tool make to run all project code.
- Generally, we want to be able to delete all files in output folders, type make all on the command line, and have all output files reappear just as they were before.

Build tool - `make`

Build tools generally help run your project code in the correct order based on a set of rules that specify output targets and inputs, including code files. Some examples of more advanced build tools include snakemake, and cmake. I have generally found that these modern build tools are unavailable on the secure servers hosted at Statistics Denmark, where a substantial part of my project code resides.
We, therefore, use the slightly more archaic make.
Documentation for make is available here.
Helpful introductions to make for data analysis
- Automation and Make by Software Carpentry.
- Makefiles for R/Latex projects by Rob Hyndman.
- Minimal make by Karl Broman. Runs a couple of R scripts and creates a latex compiled PDF paper with the resulting figures.
- GNU Make for Reproducible Data Analysis by Zachary Jones.
An important feature of make is that it compiles or runs project code based on a general recipe, the makefile. The makefile consists of targets, dependencies, and commands, which together defines rules. A makefile can contain multiple rules. These rules can be linked, for example, if a rule uses the target of another rule as a dependency.
- Target: The output file or goal you want to achieve. E.g., builddata/out/clean_data.parquet
- Dependencies: Files or targets that must be up-to-date before executing the target's commands. This will typically include the code file you want to run, and the data it uses. E.g., builddata/code/clean_data.R, and buildraw/out/rawdata.csv.
- Command: The command(s) that make will execute to create or update the target. make basically runs from the command line meaning that you must either use a command that is available from the CLI or specify the full path to the program. If, for example, you want to run an r script, you can use the CLI command rscript when this is installed.

#target: dependencies
#	command

builddata/out/clean_data.parquet: builddata/code/clean_data.R buildraw/out/rawdata.csv
	rscript builddata/code/clean_data.R

Make make run your command
1. Navigate a command line tool such as cmd to the project folder where the makefile is located.
2. In the command line, type make target, where target is the file you want to build. E.g., make builddata/out/clean_data.parquet.
How make runs your command:
- When you tell make to create a target, make first checks the dependencies for that rule. If a dependency is itself a target from another rule, makemoves back to this previous rule. This continuous until make finds the antecedent dependencies.
- make then checks the timestamps of antecedent dependencies. If the dependencies have not changed since the target was last created, make won't re-run the commands for that target. If the target does not exist, the command is always run.
- make will the move through the linked set of dependencies, running the commands from rules where dependencies have changed since the target was last created.
- make will let you know before it tries to run if a dependency does not exist and there exists no rule to create it.
Installation:
- make is usually pre-installed on Unix-based systems (like macOS and Linux). For Windows, you can install it through tools like Cygwin, GNUWin32, or rtools.
Important syntax notes:
- Indentation: Make sure to use a tab, not spaces, for indentation in the Makefile.
- Multiple Languages: If your workflow involves Stata or SAS scripts, you can include them in the Makefile just like R scripts. For instance, stata -b do my_analysis.do for a Stata script.
Line splitting with many dependencies
- If you file contains many dependencies it can be useful to split the dependency list over multiple lines. You can do this by writing \ and continuing the content on the next line after an indentation. If you include anything, include a space, after the backslash, make will throw an error.

builddata/out/clean_data.parquet: \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	rscript builddata/code/clean_data.R

Many targets
- A file may create multiple outputs, such as regression tables. To specify this, you simple add all targets to the left of :

builddata/out/clean_data1.parquet builddata/out/clean_data2.parquet : \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	rscript builddata/code/clean_data.R

Automatic variables
- make allows you to use automatic variables in rules.
  - $@: The name of the target
  - $<: The name of the first prerequisite
  - $^: The names of all the prerequisites, with spaces between them

builddata/out/clean_data.parquet: \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	rscript $<

Creating variables
- Creating variables can be useful, for example, for creating build rules or listing all targets you want to build.
- Variables can be assigned with :=. E.g., var := name1
- Variables can be called using $(variablename)
- You can append to a variable by adding $(variablename) on the right hand side of the assignment name; var := $(var) name2.

R := C:\Users\bxn825\scoop\shims\rscript.exe

builddata/out/clean_data.parquet: \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	$(R) $<

Creating build rules with many targets (phony targets)
- The rule all: $(targets) is often used to run all rules you want in a project, when the variable $(targets) contains a set of targets in your project.
- You can use .PHONY to explicitly declare all (and other non-file targets) as a phony target. This tells make that this target isn't a file but rather a label for a recipe to be executed.

R := C:\Users\bxn825\scoop\shims\rscript.exe

$(targets) := builddata/out/clean_data.parquet
builddata/out/clean_data.parquet: \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	$(R) $<

$(targets) := $(targets) builddata/out/clean_data.parquet 
builddata/out/clean_data.parquet: \
	builddata/code/clean_data.R \
	buildraw/out/rawdata.csv
	$(R) $<

.PHONY: all 

all: $(targets)

Subfolders

When projects contain subfolders for distinct tasks, each subfolder should have a makefile. The main project folder makefile should then tell make to enter the subfolder and run the makefile there. We store makefile settings that are common for the project in makefile.env.

Assume the following folder structure and note that we have a makefile in each subfolder executable code.

rawextract/ 
	dst/
		dat1.sas7bdat
		dat2.sas7bdat
consistency_simulation/
	code/ 
		simulate_estimator_consistency.R 
	log/
    temp/
	out/
		simulate_estimator_consistency_distribution.PDF
    makefile
builddata/
    code/ 
        clean_dat1.R 
    log/ 
    out/
        clean_dat1.parquet
    makefile
makefile
makefile.env

Content of the settings makefile.env

R := C:\Users\bxn825\scoop\shims\rscript.exe --quiet $< 1> log/$(basename $(notdir $<)).log 2>&1

Content of the main makefile

include makefile.env

.PHONY: all builddata consistency_simulation

all: builddata consistency_simulation 

builddata: 
    @echo entering builddata folder...
    $MAKE$ -C builddata 

consistency_simulation: 
    @echo entering consistency_simulation folder...
    $MAKE$ -C consistency_simulation

Content of the makefile in builddata

include ../makefile.env

# targets 
TARGETS = 

TARGETS := $(TARGETS) out/clean_dat1.parquet
out/clean_dat1.parquet: \
    code/clean_dat1.R \
    ../rawextract/out/dat1.sas7bdat
    $(R)

# run all targets 

.PHONY: all

all: $(TARGETS)

Software

Required

If you work on a windows machine, consider installing the software using scoop. We have created a PowerShell executable file that installs all required software in one go for you. To run it:

First install scoop
Download the file scoop_install_required_software.ps1 from the _setup folder in this repository.
Right-click it and choose "run with powershell".

The file then installs all the software. Remember you still have to install extension within each piece of software manually.

List of required software

Git
- Git is a command-line version control software. See more under Version Control
- Installation
  - Create a GitHub account and install the Git desktop/command line clients.
  - Installation via Scoop on Windows machines: scoop install git
VSCode
- VSCode is a general text editor you can use to edit markdown, Python, R, LaTeX, and many other file types.
- It has excellent Git version control features included
- Installation options:
  - Follow Jeppe Druedahl's guide to install VSCode.
  - Install using Scoop on a Windows machine with scoop bucket add extras; scoop install vscode (see Software/Suggestions on using ).
- You can add functionality by installing extensions. Some extensions you'll likely want to install:
  - Git History
  - GitHub Copilot (free for academic users - read more here)
  - LaTeX workshop
  - Markdown All in One
  - Python
  - Quarto
  - R (consider using this together with radian)
  - Stata Enhanced
  - vscode-pandoc
- Guides:
  - Markdown and Visual Studio Code
- We use pandoc or quarto to convert markdown files into Word, PDF (via LaTeX), beamer slides, or HTML files.
R, RStudio, rtools
- R is an open-source statistical programming language that has gotten a lot of traction in the Econometrics community. Most new developments in Econometrics are likely to arrive to R at the same time or prior to, e.g., Stata.
- Rstudio is an editor specialized for R coding. We typically use RStudio whenever we edit R code.
- rtools is a set of software tools R will need to compile some packages, including the arrow package we use for parquet format IO.
- Installation options:
  - [Installation guide for R and RStudio] by Posit (the developers behind RStudio).
  - Installation guide for rtools via CRAN.
  - Installation via Scoop on Windows machines: Scoop bucket add r-bucket https://github.com/cderv/r-bucket.git; scoop install r; scoop install rstudio; scoop install rtools
- Guides to get started using R
  - 4h R crash course by Hans H. Sievertsen. With Economics related examples
  - R for Data Science by Hadley Wickham and Garrett Grolemund.
Tinytex
- Tinytex is a lightweight LaTeX distribution that allows you to compile LaTeX documents to PDF.
- Installation options:
  - Guide from the developers.
  - Installation via Scoop on Windows machines: scoop install tinytex
Pandoc
- Pandoc is an open-source command line tool that can be used to convert between many different text formats. We will typically use it to convert between markdown and PDF/word documents.
- Installation options:
  - Guide from the developers. Installation via Scoopon Windows machines: scoop install pandoc
Quarto
- Quarto is a piece of software developed by Posit that allows you to write quarto documents containing both markdown text and integrated R, Python, or some other relevant code. The documents can be edited in RStudio or VSCode. Quarto documents are beneficial when some code requires extensive documentation.
- Installation options:
  - Guide from the developers.
  - Installation via Scoop on Windows machines: scoop install quarto
Zotero
- Zotero is an open-source reference manager.
- Allows easy export of bibtex reference libraries that can cited in markdown and latex documents.
- Installation options:
  - Guide from the developers.
  - Installation via Scoop on Windows machines: scoop install zotero
- Install connectors:
  - Installation guide from the Zotero developers.
- Install relevant extensions
  - better bibtex
    - Install using the guide at the link.
    - Set bbt's citation key formula to authEtAl + year + shorttitle(3,3) by going to preferences --> Better BibTex --> Open Better BibTex Preferences --> Citation Key, and copy the formula in.

Suggestions

Scoop
- A Powershell tool for Windows that helps you install and update software updated.
- Installation options:
  - Guide from the developers.
- Working on university-provided IT equipment can give update and installation problems if you do not have administrator rights over the computer. This can be circumvented by ensuring that all (or most) programs are installed in your own user path. Scoop does this.
- We presently use the Powershell tool scoop to manage the installation of most software on my system. scoop uses recipes created by others to install (often) the latest versions of programs.
- Installing a program with scoop is as simple as scoop install program.
- Updating a program with scoop is as simple as Scoop update program. Type scoop update * to update all installed programs.
- Scoop searches for install recipes in buckets. Buckets can be added by typing scoop bucket add ... in Powershell. Examples of useful buckets include extras, nerd-fonts, and r-bucket. You can find an example of installing and using a scoop to set up a new Windows machine at https://github.com/EriksenJ/_setup.
Obsidian
- A markdown-based digital note editor with latex compilation to PDF ready out of the box.
- Some suggested plugins
  - LaTeX Suite
  - Linter
  - Obsidian Git
  - Templater
  - Zotero Integration
- Install with scoop on windows: scoop bucket add extras; scoop install obsidian
Linh T. Tô has a great set of (free) resources on her website.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README.md

README.md

Introduction

Workflow

Tasks

Reporting and notes

References

Writing papers

Version Control

Project structure

Project storage

Directories

Data formats

R projects

Code Conventions

General

R

Larger than memory datasets

Guides to get started with R

Automation

Build tool - `make`

Subfolders

Software

Required

Suggestions

Files

README.md

Latest commit

History

README.md

File metadata and controls

Introduction

Workflow

Tasks

Reporting and notes

References

Writing papers

Version Control

Project structure

Project storage

Directories

Data formats

R projects

Code Conventions

General

R

Larger than memory datasets

Guides to get started with R

Automation

Build tool - make

Subfolders

Software

Required

Suggestions

Build tool - `make`