01.01: Description, prediction, and explanation

[malcolmbarrett]

• [x] A Second Chance to Get Causal Inference Right: A Classification of Data Science Tasks: https://amstat.tandfonline.com/doi/full/10.1080/09332480.2019.1579578
• [x] #53, data: https://osf.io/jtdaz/
• [x] to explain or predict (see below)
• [x] table two fallacy: https://academic.oup.com/aje/article/177/4/292/147738, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6103824/
• [ ] mention how techniques in prediction modeling do and don't apply

Description papers

• [ ] https://twitter.com/EpidByDesign/status/1559632956687372293
• [ ] https://twitter.com/annakporter/status/1559210713100570624
• [ ] https://twitter.com/WhitneyEpi/status/1559609155614375937
• [ ] https://twitter.com/maryaglipay/status/1559654992037961732
• [ ] https://twitter.com/Epi_D_Nique/status/1564378859205808131
• [ ] https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(09)61922-X/fulltext
• [ ] https://fn.bmj.com/content/99/5/F413
• [ ] Data journalism examples? E.g. FT COVID graphs from early in the pandemic: https://www.ft.com/content/a2901ce8-5eb7-4633-b89c-cbdf5b386938
• [ ] https://twitter.com/EpiEllie/status/1574394792414437380

Prediction papers

• [x] https://www.thelancet.com/journals/lanres/article/PIIS2213-2600(20)30559-2/fulltext
• [x] https://www.bmj.com/content/369/bmj.m1328

Causal Inference papers

• [ ] https://www.nejm.org/doi/full/10.1056/NEJMoa2211029
• [ ] https://twitter.com/_MiguelHernan/status/1559425817322700802
• [ ] https://obgyn.onlinelibrary.wiley.com/doi/10.1111/1471-0528.15659
• [ ] https://obgyn.onlinelibrary.wiley.com/doi/10.1111/1471-0528.16906
• [ ] https://www.recoverytrial.net/

Examples provided by Scott Cunningham

• [ ] https://www.sciencedirect.com/science/article/abs/pii/S0094119010000896
• [ ] https://www.scunning.com/files/cunningham-and-finlay-2012.pdf
• [ ] https://dmarkanderson.com/ECNS432_Fall2020/JHE_MethProject.pdf
• [ ] https://www.aeaweb.org/articles?id=10.1257/pol.3.3.1
• [ ] https://academic.oup.com/qje/article-abstract/133/1/407/4060075?login=false

RCTS

• [ ] Describe the ideal
• [ ] Discuss A/B tests
• [ ] https://www.nytimes.com/2022/09/24/business/linkedin-social-experiments.html

[tgerke]

Possible tidbit to include: here's a proof with simple lm showing how you can use a causally biased model to improve prediction performance (i.e. answers the question "wouldn't the best prediction model always be the precise causal model?" with the answer "no.").

Slide 10 here shows the relevant section from this paper