From Curiosity to Question

Lecture · Module 5 · Wed Sep 16

Preparing for Project 1

Modules 1 through 5 have given you the pieces of a descriptive-analysis workflow: how to move from a question, to a population, to measured variables, to tables and figures, and finally to a reproducible document that tells a clear data story.

Today, you will begin using that workflow on a topic your team chooses.

Project 1 runs on Pew Research Center’s American Trends Panel — the same survey you worked with on Monday. In the M05 lab you registered with Pew, downloaded Wave 163 as a .sav, read it into R, and built a Table 1 from it. You already know what a wave, a codebook, and a questionnaire are, so none of that is what today is about.

What changes is that the wave was chosen for you then, and you choose it now. The ATP has run more than 160 waves — politics, religion, technology, work, family, immigration, science — and your first job is finding the one that asked about something your team actually wants to describe.

By the end of today’s session, your Project 1 team will be formed. You will choose a topic your team cares about, find an ATP wave that contains useful variables, and draft a research question that can guide your analysis.

Your question does not need to be perfect today. It just needs to be a strong starting point: descriptive, answerable from the wave, narrow enough to complete in four weeks — and worth someone’s time to read.

What today feeds

Project 1 asks your team to post documentation/project-plan.md to your repository. It has eleven sections. Today fills in six of them.

Section	Filled today?
Team	✅
How we will work together	✅
Topic	✅
Research question	✅ draft
Intended reader	—
ATP wave	✅ candidate
Focal variables	—
Planned comparison	✅ draft
Survey weight	—
Measurement question	—
Feasibility	—

Everything today happens on paper. Your repository does not exist yet. You will transcribe this into the plan file once it does.

Before you look · what makes an idea interesting?

How to recognize an idea worth describing

Interesting descriptive questions usually begin with one of these sparks:

Spark	What to look for	Example question shape
A surprising number	A statistic that makes you pause	How common is ___ among ___?
A group difference	A pattern that seems to vary by age, gender, education, party, religion, income, or region	How does ___ differ by ___?
A public argument	A claim people make confidently but rarely put numbers behind	What share of ___ believe or report ___?
A population you care about	A group connected to your field, work, or research interests	Among ___, how common is ___?
A mismatch	A gap between what people say, believe, experience, or do	How do ___ and ___ line up?
A measurement question	A concept that sounds simple until you ask how it was measured	How did Pew measure ___, and what does that reveal?

A good Project 1 idea does not need to be dramatic. It needs to be describable — a pattern that could become a table, a figure, and an insightful commentary.

Part 1 · Find something that interests you

In this first part, you are going shopping for an idea. The goal is not to find the perfect project in twenty minutes. The goal is to find a topic that makes you say:

“There is something here worth describing carefully.”

Work alone. Two windows open:

Pew’s research topics — start here. Thirteen broad categories: politics, religion, science, internet and technology, family, economy and work, immigration, news habits, and more. Click one and you get its reports; open a report and it names the wave behind it. (Already have a word in mind? The full A–Z topic list has hundreds of narrower entries.)
the ATP dataset index — the waves themselves, for once you know which one you are after

Go topic → report → dataset, in that order. Reports are written to be interesting; codebooks are not. Browsing a codebook cold is the slowest way to find something you care about, because it shows you variable names before it shows you why anyone bothered to ask.

Then apply the “I wonder…” test. When a Pew finding catches you, pause and finish one of these sentences:

I wonder how common this is among…
I wonder whether this differs by…
I wonder whether the same pattern appears for…
I wonder how this is measured…
I wonder what the distribution looks like…
I wonder whether people who say ___ also say ___…

Those are the beginnings of descriptive research questions. If you cannot finish any of the six, the finding is interesting to read but is not yet yours to describe.

Write down two or three possibilities

For each possible idea, record:

Topic	The substantive issue, in a phrase
What made it interesting	One sentence — the finding, contrast, or claim that caught you
Possible population	Who the question might be about
Possible comparison	The group difference, category, or pattern you might describe
Wave	Wave number, or the report title if you have not found the wave yet

A topic without a wave attached does not count.

Don’t evaluate everything yet. Don’t worry whether the variables are perfect. We’re just collecting promising ideas.

If nothing is catching your attention

Completely normal. Browsing can feel overwhelming because everything is potentially interesting.

Try one of these routes.

Route 1 · Start from a population

Pick a group you care about, then look for what Pew asked them.

Examples:

young adults
parents
rural residents
immigrants
religiously unaffiliated adults
people with lower trust in institutions
adults who use social media for news

Then ask:

What would be useful to know about this group?

Route 2 · Start from a comparison

Pick a difference that might matter.

Examples:

younger vs. older adults
parents vs. non-parents
Democrats vs. Republicans
college graduates vs. non-college graduates
frequent vs. infrequent social media users
religiously affiliated vs. religiously unaffiliated adults

Then ask:

What attitude, behavior, or experience might differ across these groups?

Route 3 · Start from a public claim

Think of something people say confidently.

Examples:

“Young adults do not trust institutions.”
“People are burned out by the news.”
“Social media makes people more politically divided.”
“Parents are especially worried about technology.”
“People do not trust science anymore.”

Then turn the claim into description:

What proportion of people report this?
Does it differ across groups?
How was the idea measured?

Route 4 · Start from a word

Skim Pew’s full A–Z topic list for a word that makes you curious.

Look for words like:

worry
trust
confidence
approval
belonging
discrimination
loneliness
news
social media
religion
family
work
safety
science

A concept that makes you curious can become the center of a question — and the topic index will show you which reports, and so which waves, actually measured it.

Part 2 · Put it on the board

Round-robin

Give us your best topic:

“I’d like to look at ____, because ____. It’s in Wave ____.”

Topics go on the board as they are said. Hopefully clusters will start to appear on their own.

Your job while others are talking: make note of the topics that interest you.

Part 3 · Form teams

Move to your cluster

Stand up and go to the topic on the board that you most want to work on.

Teams of 3. Twelve of us means four teams. The project permits teams of 2 or 3, so if a cluster comes out at two people and everyone is content, that is a real team.

When your team has settled, agree on four things. All of it goes into documentation/project-plan.md, and none of it takes long.

One shared topic. One — not a merger of three interests. A question broad enough to hold everyone’s topic is too broad to answer in four weeks.
Who keeps today’s paper plan, and who transcribes it. One person holds the paper so it does not evaporate at the end of class; someone — the same person or another — types it into documentation/project-plan.md once the repository exists. Name both.
Who owns the repository. One person unzips the starter, makes it an RStudio project, publishes it to GitHub as private, and invites the teammates plus both instructors. This is an administrative job only — it carries no extra say over the analysis. The project setup guide is their walkthrough.
How you will work together. The plan file asks for two of these by name, and the third saves you a week of merge conflicts:
- When you will meet. A standing time beats “we’ll message each other.”
- How long you wait before checking in if someone goes quiet — and who tells the instructor.
- A first pass at who writes which section. The brief has a handful — the question, measurement, the analytic sample, Table 1, each figure, what the findings mean — and each one gets a primary editor, the person who makes the keystrokes there. That is what keeps two of you out of the same lines of the same file. Provisional is fine today; it lives in the plan file’s Team section and you can revise it. The rubric works a three-person split all the way through — see the worked example, which gives each of you something code-heavy, something writing-heavy, and one documentation file. Start from it rather than inventing your own.

Settle all of it now, while it is still hypothetical — that is exactly why the plan file asks.

Two rules about the split

Divide by section, not by competence. The temptation is to hand the code to whoever is quickest at R and the writing to whoever is quickest at prose. Do not: everyone needs hands-on time with the data and the code, and the rubric scores that individually.

Primary editor is not sole expert. It means who types there, not who understands it. At least one teammate reads every section, and in the presentation any of you can be asked about any part of the analysis.

Part 4 · Draft the question

Now the real work. A topic is not a question. Social media and teens is a topic. It cannot be answered, confirmed, or plotted.

Remember the three tasks

In our first meeting we separated description, prediction, and causal inference, and said the wording of a question gives you a fast first guess but not the answer.

That distinction now has consequences. Project 1 is a description project. It asks what the world looks like, for a named population, on measured variables.

It is not a prediction project and not a causal one. The project also asks you not to report confidence intervals, margins of error, or p-values — that machinery arrives in Modules 6 through 9, and Project 2 is where you use it.

So the question you write today has to be one that description can actually answer.

Descriptive question shapes that work well for Project 1

If your team is stuck, start with one of these shapes.

Shape	Template
Prevalence	Among [population], how common is [attitude, behavior, or experience]?
Group difference	Among [population], how does [outcome] differ by [grouping variable]?
Distribution	Among [population], what does the distribution of [variable] look like?
Co-occurrence	Among [population], how often do [response A] and [response B] appear together?
Profile	Among people who [report X], what else do they report?
Comparison within a topic	Which [concern, source, reason, or response] is most common among [population]?

For Project 1, the strongest questions are usually simple:

Among a named population, how does one measured thing vary across one meaningful comparison?

The fourth check · is it worth asking?

Descriptive, answerable, narrow — those three keep a question feasible. There is a fourth check, and on the rubric it carries more weight than any other single thing about your project: roughly 8 of the 25 project-specific points ride on the question itself.

A question can pass all three checks and still not earn a brief. How common is social-media use, by age? fits a shape above, is perfectly answerable, and changes nobody’s understanding when they read the answer.

So put your draft through one more test:

If I already knew the answer, would I bother reading this?

If the honest answer is no, you do not need a harder question — you need a better-chosen one. Three moves usually do it:

Narrow the population to the group the question is actually about — not all adults, but adults who get news on social media.
Ask about something contested rather than something everyone already predicts.
Compare two measures of the same idea instead of describing one — that gap is often where the interesting finding lives.

None of these makes the analysis harder. They are all M01–M05 work.

Team drafting

Step 1 · Write a draft of your research question.

Step 2 · Jot down the following initial elements of your plan.

Team
Topic
Research question
ATP wave
Planned comparison
The thing we are least sure of

What a question looks like after all four checks

Too broad:
How does social media use vary among young adults

This is not ready for Project 1.

“Young adults” needs a clearer population.
There is no planned comparison.

Better:
Among U.S. adults ages 18–29, how does self-reported social media use differ by employment status?

This version is not more impressive. It is more answerable.

It is descriptive.
It names a population.
It names measurable variables.
It has one comparison.
It could become a table, a figure, and a brief commentary.

Stronger still: that draft clears the first three checks, but how much 18–29s use social media by employment status is close to something we could already guess. Keep every tool the same and move the outcome to something contested:

Among U.S. adults ages 18–29 who use social media daily, how does the share who say it makes them feel more connected differ by employment status?

Same population logic, same single comparison, same M01–M05 analysis — but now the answer is genuinely uncertain before you run it.

For this project, that is the goal: not the biggest question, but the clearest one you can answer honestly and one a reader would actually want the answer to.

From here, you work as a team

Your team exists as of today, and Project 1 is now your instruction set — not this page. Nothing else happens in class before the proposal is due; the work below is yours to schedule between now and then, at the meeting rhythm you just agreed on.

What your team must finish by Friday, September 25, 5 PM

That is the proposal deadline, and it closes Week 1 of the project — the first row of the Project 1 calendar. Four things, in this order:

Set up the repository. Your repo owner unzips the starter, makes it an RStudio project, publishes it to GitHub as private, and invites the teammates and both instructors. The project setup guide is the walkthrough. Do this first — the other three need it to exist.
Transcribe today’s paper plan into documentation/project-plan.md, then fill in the five sections we did not reach today: intended reader, focal variables, survey weight, the measurement question you already have, and feasibility.
Confirm your wave actually contains your variables. Open the codebook and find each one by name. This is the single highest-value hour your team will spend this week — a variable you assumed was there is the most common way a team loses Week 2.
Push it. The proposal is the pushed file, so none of the above counts until it is on GitHub with both instructors added as collaborators. A repository we cannot open is not a submission.

Note what is not on this list: the data. You do not need to download the .sav by the proposal deadline — that is Week 2 work. Week 1 is a question, a wave, and a working repository.

Full requirements are on the Project 1 page.

Your question may change. That is not unexpected, and the plan file has a Changes to this plan section precisely because it should. What matters is that it changes deliberately, from a starting point you chose, rather than drifting because you never fixed one.