5.5 Estimation

  • Simply compare outcome averages in treatment and control group
    • e.g., using t-tests or simple linear regression model (OLS)
  • Always possible to apply further operations on the data (e.g., take subsample etc.)
  • Sometimes exploring treatment heterogeneity is warranted
    • Subgroups could be affected differently
    • e.g., using novel machine learning methods (cf. Bauer and Clemm von Hohenberg 2020)

References

Bauer, Paul C, and Bernhard Clemm von Hohenberg. 2020. “Believing and Sharing Information by Fake Sources: An Experiment.”