4.16 Why moving from ITE to ATE? (Wikipedia)

• ITE is unobservable
• Average TE = Diff. in averages
• ATE can be estimated relying on less daring assumptions (Holland 1986, 948f)
  • e.g., (proper) random assignment (should) partly satisfy them (“statistical solution”)
• Estimating Population ATE (PATE) from Sample ATE (SATE) requires random sample (Wikipedia, Holland 1986, @Imai2008–hj)

References

Holland, Paul W. 1986. “Statistics and Causal Inference.” J. Am. Stat. Assoc. 81 (396): 945–60.