US Preventive Services Task Force (USPSTF) (2016). Final recommendation statement: Breast cancer: Screening. Available at: https://www.uspreventiveservicestaskforce.org/Page/Document/RecommendationStatementFinal/breast-cancer-screening1 (updated: January 2016).

Wikipedia. (2018). Graphoid. Available at: https://en.wikipedia.org/wiki/Graphoid (last edited: January 8, 2018).

Wiegerinck, W., Burgers, W., and Kappen, B. (2013). Bayesian networks, introduction and practical applications. In Handbook on Neural Information Processing (M. Bianchini, M. Maggini, and L. C. Jain, eds.). Intelligent Systems Reference Library (Book 49). Springer, Berlin, Germany, 401–431.

Глава 4. Осложнители и наоборот: как убить прячущуюся переменную

Глава 4. Осложнители и наоборот: как убить прячущуюся переменную

Annotated Bibliography

The story of Daniel has frequently been cited as the first controlled trial; see, for example, Lilienfeld (1982) or Stigler (2016). The results of the Honolulu walking study were reported in Hakim (1998).

Fisher Box’s lengthy quote about “the skillful interrogation of Nature” comes from her excellent biography of her father (Box, 1978, Chapter 6). Fisher, too, wrote about experiments as a dialogue with Nature; see Stigler (2016). Thus I believe we can think of her quote as nearly coming from the patriarch himself, only more beautifully expressed.

It is fascinating to read Weinberg’s papers on confounding (Weinberg, 1993; Howards et al., 2012) back-to-back. They are like two snapshots of the history of confounding, one taken just before causal diagrams became widespread and the second taken twenty years later, revisiting the same examples using causal diagrams. Forbes’s complicated diagram of the causal network for asthma and smoking can be found in Williamson et al. (2014).

Morabia’s “classic epidemiological definition of confounding” can be found in Morabia (2011). The quotes from David Cox come from Cox (1992, pp. 66–67). Other good sources on the history of confounding are Greenland and Robins (2009) and Wikipedia (2016).

The back-door criterion for eliminating confounding bias, together with its adjustment formula, were introduced in Pearl (1993). Its impact on epidemiology can be seen through Greenland, Pearl, and Robins (1999). Extensions to sequential interventions and other nuances are developed in Pearl (2000, 2009) and more gently described in Pearl, Glymour, and Jewell (2016). Software for computing causal effects using do-calculus is available in Tikka and Karvanen (2017).