Interpreting results

Lecture 17

Aidan Combs

Duke University
SOCIOL 333 - Summer Term 1 2023

2023-06-15

Logistics

• Four classes left. No class Monday. We’re almost there! :)

Wrapping up the project

• Project component 3: submit for grading by next Tuesday (June 20)

• Final paper

  • use (revised!) elements of your proposal, descriptive stats, and results to turn your question and analysis into a (concise) paper.
  • More detailed instructions and example will be posted soon
  • Due Wednesday June 28 11:59pm

• Final presentation

  • Short! ~5-10 minutes
  • Briefly introduce your question, say a bit about why it’s interesting, show your results, and talk about what you conclude from them.
  • Most on Monday June 26 (last day of class)

Final presentation logistics

• Looking for a few volunteers to go on Thursday June 22 instead!

  • Benefits…
  • More time afterwards to focus on your final paper (finish up the semester early?)
  • No stress while you watch day 2
  • Everyone gets more flexibility on time
  • What’s your favorite kind of donut? :)

Today

• Interpreting results!

Yesterday’s exercise

The (approximate) data analysis process

• Determine topic ✓
• Find data; learn what observations and variables are available ✓
• Write research question ✓
• Describe distributions of relevant variables ✓
• Prepare data frame for analysis ✓
• Describe relationships between variables ✓
• Perform statistical tests ✓
• Communicate results

So you have a p value…

• Rejecting/failing to reject hypotheses is useful
• But it’s also statistics-speak, not effective communication

What do you want to be able to say?

• Something about what the world is like!
• Translate back from hypothesis test language—zoom back out. If you were describing your results to a friend, what would you say you learned about the world?

What can’t we say?

• Knowing the limitations of your conclusions is super important!
• Limitations can arise from several places.

Sampling and generalizability

• Who can your results speak to?
• Is your sample big enough?
• Was it selected randomly?

Accuracy

• Are people likely to give you accurate answers?

• Unintentional inaccuracies

  • Desirability biases
  • Memory problems

• Misrepresenting/lying

  • Is your data coming from something or someone who has a stake in it?

• Saying vs doing

  • What people say is often different than what they do
  • Stated ideals often don’t match up with actions
  • “I believe in equality but I oppose all the policies that would create it”
  • Did your data collect information on saying, or on doing?
  • Avoid drawing conclusions about what people would do when the data you have is about what they say

Operationalization

• Are the questions and categories what you would like them to be?

• Do we lose information through bad category choice or bad documentation?

  • For example: bad but common operationalizations

    • Race: white or nonwhite
    • Gender: male or female

Causality

• Association vs causation: What’s the difference?

• Association: two things are related
• Causation: one thing drives another thing

• It’s nice to be able to say something about causation, but it’s hard!
• Hypothesis tests speak to association. They sometimes speak to causation as well—it all depends on the set up of your study!

Exercise Q1

• Does this provide evidence that using the internet helps people live longer? Why or why not?
• Write down a plausible explanation for this relationship.

Reading for Tuesday

• There’s an article linked on the course calendar. Mess around with the interactive element!

Workshopping results

• Teams

• Same as last time: you have access to each other’s repos on GitHub. Leave your feedback as an issue. Questions to address are at the top of the instructions page on the website (link).