Wright, S. (1983). On “Path analysis in genetic epidemiology: A critique.” American Journal of Human Genetics 35: 757–768.

Глава 3. От доказательств к причинам. Преподобный Байес знакомится с мистером Холмсом

Глава 3. От доказательств к причинам. Преподобный Байес знакомится с мистером Холмсом

Annotated Bibliography

Elementary introductions to Bayes’s rule and Bayesian thinking can be found in Lindley (2014) and Pearl, Glymour, and Jewell (2016).

Debates with competing representations of uncertainty are presented in Pearl (1988); see also the extensive list of references given there.

Our mammogram data are based primarily on information from the Breast Cancer Surveillance Consortium (BCSC, 2009) and US Preventive Services Task Force (USPSTF, 2016) and are presented for instructional purposes only.

“Bayesian networks” received their name in 1985 (Pearl, 1985) and were first presented as a model of self-activated memory. Applications to expert systems followed the development of belief updating algorithms for loopy networks (Pearl, 1986; Lauritzen and Spiegelhalter, 1988).

The concept of d-separation, which connects path blocking in a diagram to dependencies in the data, has its roots in the theory of graphoids (Pearl and Paz, 1985). The theory unveils the common properties of graphs (hence the name) and probabilities and explains why these two seemingly alien mathematical objects can support one another in so many ways. See also “Graphoid,” Wikipedia.

The amusing example of the bag on the airline flight can be found in Conrady and Jouffe (2015, Chapter 4).

The Malaysia Airlines Flight 17 disaster was well covered in the media; see Clark and Kramer (October 14, 2015) for an update on the investigation a year after the incident. Wiegerinck, Burgers, and Kappen (2013) describes how Bonaparte works. Further details on the identification of Flight 17 victims, including the pedigree shown in Figure 3.7, came from personal correspondence from W. Burgers to D. Mackenzie (August 24, 2016) and from a phone interview with W. Burgers and B. Kappen by D. Mackenzie (August 23, 2016).

The complex and fascinating story of turbo and low-density parity-check codes has not been told in a truly layman-friendly form, but good starting points are Costello and Forney (2007) and Hardesty (2010a, 2010b). The crucial realization that turbo codes work by the belief propagation algorithm stems from McEliece, David, and Cheng (1998). Efficient codes continue to be a battleground for wireless communications; Carlton (2016) takes a look at the current contenders for “5G” phones (due out in the 2020s).