Onboarding Homework

Author

Affiliation

Your Name Here

Old Dominion University

Before You Start

This assignment has two jobs. First, confirm that your toolkit works: R, RStudio, your course R Project, your CLAUDE.md, Claude Code, and Quarto rendering. Second, get a first look at the central question of this course: when can a difference between groups be read as an effect?

Setup, in order:

Create your course folder and put the course .Rproj and CLAUDE.md files in it.
Put this file in your course folder (or one subfolder deep, such as a homework folder inside it, but no deeper).
Open the project in RStudio by double-clicking the .Rproj file, then open this file.
Install the dagitty package (install.packages("dagitty") in the Console). You only need to do this once.
Change the author line at the top of this file to your name.
Click Render right now, before doing anything else. If a document appears, you’re in business. Render early and often.

How to use AI: You may (and should!) use AI tools to help you throughout, but what you submit is your responsibility, not the AI’s. Read what it gives you, check it against the course material, and change whatever you would not stand behind. Never submit something you don’t understand, and make sure that what you submit is written in your own voice.

What to submit: Submit the rendered .html file on Canvas. There is no video. Your report is graded on completeness and correctness, not the course rubric.

The data

You will work with two datasets about race and the labor market. Use only these.

Resume audit (data; documentation; published paper): in 2001 and 2002, two economists sent thousands of fictitious resumes to real job ads in Boston and Chicago. Names were randomly assigned (some sounded stereotypically white, others stereotypically African-American), and qualifications were randomly varied. The outcome is whether the employer called back.
ACS (data): the 2011 American Community Survey, a Census Bureau survey of roughly 1% of U.S. households. Each row is an adult aged 25 to 54 living in Boston or Chicago who is either non-Hispanic white or non-Hispanic Black. The outcome is employed.

How the ACS works: the ACS surveys a sample of households, and not everyone is equally likely to be sampled. Each row therefore carries a weight, pwgtp, which is the number of people in the population that row represents. Unweighted summaries describe the sample; weighted summaries describe the population. Think about when each matters.

Question 1: Setup Check

(5 Points) The chunk below checks your setup. Do not edit it; just make sure it runs when you render. Full credit requires the output to show .Rproj found: TRUE, CLAUDE.md course block: INTACT, and dagitty installed: TRUE.

Question 2: Screenshot

(5 Points) Take a screenshot of your RStudio window that shows all three of the following:¹

Your project name in the top-right corner of RStudio (this proves the .Rproj is actually open).
The Terminal tab (bottom-left) with Claude Code running in it.
In that terminal, Claude Code’s answer to this exact prompt: “In one sentence: what folder are you working in, and what files do you see?”

Embed the screenshot below by replacing the file name with your own:

Question 3: Report

(30 Points) Write a short report on the two datasets, in your own words, using the vocabulary of this course. Use these sections, in this order.

a. The data (6)

Describe each dataset: how was it collected, what does one row represent, and how do the two differ?

b. The effect of race (8)

Use both datasets to estimate the effect of race on the outcome. For each, estimate an unconditional version and a conditional version (with controls).

c. Causal or not? (12)

Why did you choose the controls that you did?
Which of these estimates (if either) delivers a causal effect? Why or why not? Use DAGs or DAG language (see the Module 1 Identification notes).
What else could produce the gap?

d. Process Reflection (4)

In your own words (Claude Code is not allowed to write this, and it knows that): how did you use Claude on this homework, where did its first attempt fall short, and what did you change or reject, and why?

Footnotes

On Windows: Win + Shift + S. On Mac: Cmd + Shift + 4. Save the image file in the same folder as this document.↩︎

--- title: "Onboarding Homework" author: "Your Name Here" institute: "Old Dominion University" format: html: theme: united code-tools: true code-fold: true code-summary: "Code" code-copy: hover link-external-newwindow: true tbl-cap-location: top fig-cap-location: bottom self-contained: true editor: source --- ```{r setup, include=FALSE} # INCLUDE THIS AT THE START OF EACH ASSIGNMENT # DO NOT EDIT THIS knitr::opts_chunk$set(fig.align = 'center') knitr::opts_chunk$set(out.width = '90%') knitr::opts_chunk$set(results = 'hold') knitr::opts_chunk$set(fig.show = 'hold') knitr::opts_chunk$set(dev='svg') knitr::opts_chunk$set(error = TRUE) par(mar = c(4.1, 4.1, 1.1, 4.1)) ``` ## Before You Start This assignment has two jobs. First, confirm that your toolkit works: R, RStudio, your course R Project, your `CLAUDE.md`, Claude Code, and Quarto rendering. Second, get a first look at the central question of this course: when can a difference between groups be read as an effect? Setup, in order: 1. Create your course folder and put the course `.Rproj` and `CLAUDE.md` files in it. 2. Put **this file** in your course folder (or one subfolder deep, such as a `homework` folder inside it, but no deeper). 3. Open the project in RStudio by double-clicking the `.Rproj` file, then open this file. 4. Install the `dagitty` package (`install.packages("dagitty")` in the Console). You only need to do this once. 5. Change the `author` line at the top of this file to your name. 6. Click **Render** *right now*, before doing anything else. If a document appears, you're in business. Render early and often. **How to use AI**: You may (and should!) use AI tools to help you throughout, but what you submit is your responsibility, not the AI's. Read what it gives you, check it against the course material, and change whatever you would not stand behind. Never submit something you don't understand, and make sure that what you submit is written in your own voice. **What to submit**: Submit the rendered `.html` file on Canvas. There is no video. Your report is graded on completeness and correctness, not the [course rubric](https://alexcardazzi.github.io/econ708a/rubric.html). --- ### The data You will work with two datasets about race and the labor market. Use only these. - **Resume audit** ([data](https://vincentarelbundock.github.io/Rdatasets/csv/AER/ResumeNames.csv); [documentation](https://vincentarelbundock.github.io/Rdatasets/doc/AER/ResumeNames.html); [published paper](https://www.aeaweb.org/articles?id=10.1257/0002828042002561)): in 2001 and 2002, two economists sent thousands of fictitious resumes to real job ads in Boston and Chicago. Names were randomly assigned (some sounded stereotypically white, others stereotypically African-American), and qualifications were randomly varied. The outcome is whether the employer called back. - **ACS** ([data](https://alexcardazzi.github.io/econ708a/data/acs_2011_boston_chicago.csv)): the 2011 American Community Survey, a Census Bureau survey of roughly 1% of U.S. households. Each row is an adult aged 25 to 54 living in Boston or Chicago who is either non-Hispanic white or non-Hispanic Black. The outcome is `employed`. **How the ACS works**: the ACS surveys a sample of households, and not everyone is equally likely to be sampled. Each row therefore carries a weight, `pwgtp`, which is the number of people in the population that row represents. Unweighted summaries describe the sample; weighted summaries describe the population. Think about when each matters. --- ## Question 1: Setup Check **(5 Points)** The chunk below checks your setup. Do not edit it; just make sure it runs when you render. Full credit requires the output to show `.Rproj found: TRUE`, `CLAUDE.md course block: INTACT`, and `dagitty installed: TRUE`. ## Question 2: Screenshot **(5 Points)** Take a screenshot of your RStudio window that shows all three of the following:^[On Windows: `Win + Shift + S`. On Mac: `Cmd + Shift + 4`. Save the image file in the same folder as this document.] - Your project name in the **top-right corner** of RStudio (this proves the `.Rproj` is actually open). - The **Terminal tab** (bottom-left) with Claude Code running in it. - In that terminal, Claude Code's answer to this exact prompt: *"In one sentence: what folder are you working in, and what files do you see?"* Embed the screenshot below by replacing the file name with your own: ![My RStudio setup](your_screenshot_file_name.png) ## Question 3: Report **(30 Points)** Write a short report on the two datasets, in your own words, using the vocabulary of this course. Use these sections, in this order. #### a. The data (6) - Describe each dataset: how was it collected, what does one row represent, and how do the two differ? #### b. The effect of race (8) - Use both datasets to estimate the effect of race on the outcome. For each, estimate an unconditional version and a conditional version (with controls). #### c. Causal or not? (12) - Why did you choose the controls that you did? - Which of these estimates (if either) delivers a causal effect? Why or why not? Use DAGs or DAG language (see the Module 1 Identification notes). - What else could produce the gap? #### d. Process Reflection (4) - In your own words (Claude Code is not allowed to write this, and it knows that): how did you use Claude on this homework, where did its first attempt fall short, and what did you change or reject, and why?