Day 12

Dad Joke

Did you hear about the mathematician who was afraid of negative numbers?

He would stop at nothing to avoid them


  • IRB Discussion
  • Ask for help
  • Rough draft for peer feedback on December 2?

Paper Discussion

DellaPosta, D., Shi, Y., & Macy, M. (2015). Why Do Liberals Drink Lattes? American Journal of Sociology, 120(5), 1473–1511.

Why isn’t politics just about policies?

  • We all have stereotypes about what conservatives and liberals are like
  • What do conservatives wear?
  • What cars do liberals drive?
  • In some cases, these are more correlated with party than policy issues!

Explaining correlations rely on a search for explanations

  • E.g., liberals were more likely to believe in astrology
    • Why? Maybe because it is seen as superstitious/sacrilegious?
  • But a similar just-so story could be told if the correlation went the other way
    • Conservatives have a more magical / less scientific worldview

What’s the answer? People are part of networks!

  • “Alignment [is an] emergent property of a self-organizing cultural landscape”
  • We want to be more like the people who we see as our group
    • Social influence: we change our beliefs/opinions/behaviors
    • Homophily: we change who we are connected to and connect to those more like us
  • Result is networks that are autocorrelated

Simulation Model

  • Agents are given 25 random “opinions” of 1/0
  • The (observed - expected) Euclidian distance is the weight between them
  • Agents probabilisitically adopt beliefs of neighbors based on similarity

Even when opinions are correlated with demographics, it could be network autocorrelation that causes both

Agent-based modeling is neat

  • Using computational agents
  • Test theory
    • Can agents acting according to theory produce what we see empirically?
  • Tool for thinking
    • How might we expect changes to relate to higher level dynamics?

Principles for code organization

  • 3 directories
    • code
    • data
    • paper

Reproducible workflows and best practices

  • The idea is to be able to fix any step in the workflow and have the rest run with minimal problems
    • Data updates
    • Bugs
    • Changes to measures or questions
  • Save raw data
  • Run code in order
  • Write comments! (or text in Jupyter)

HW Review

  • Wrong branch of the class cloned - oops!