Lecture 1 Changes

[mkiang]

• [x] Still need example of good CI
• [x] Missing some sources
• [x] > One suggestion I have is to do a review at the transition between our motivation and the causal framework. Something short to summarize and reiterate our points. Also, I'll help add 1-2 more examples as I review literature in the next week. - @writezhe
• [x] > Examples of counterfactuals in various settings (some easy, some hard) twins, change to Amazon business model, consumer demand shock, policy change, election politics, medicine, change to the transportation system. - @writezhe
• [x] h tags might need to be more consistent — unclear.
• [x] Add h titles to beginning slides
• [x] Not all ? in example table data are .red[]
• [x] Additional reading probably needs additional readings
• [x] Not all slides have presenter notes — Add them before I forget what I was doing with them

[mkiang]

Lecture 1 is ready for review: https://mkiang.github.io/intro-ci-shortcourse/slides/part-01-intro/

Just go to that page and hit "P" key to see presenter notes. The CSS layout needs to be modified but am looking for feedback on typos and or content and or ideas.

[mkiang]

• [x] Add consistency to the list of assumptions

[MJAlexander]

hey @mkiang Lecture 1 looks really good (also xaringan super fancy!). A few minor comments:

• [x] Stress difference between causal and predictive interpretations. Presumably data science students are used to thinking about problems mostly in terms of prediction. So maybe discuss slide 17 (definition of causal inference) wrt how it differs from predictive inference (e.g. comparisons between units vs comparisons of different treatments on same units).
• [x] RE: other examples: Jas Sehkon has a lot of cool examples in poli sci. But that said probably best if you talk about papers you know more
• [x] Maybe add a slide after slide 18 introducing Rubin causal framework in a couple of sentences, rather than just jumping in
• [x] slide 24: Assumptions for/of what? Assumptions of RCM sufficient to identify ATE?
• [x] for exchangeability statement: want double line independence symbol instead of $\perp$ but unsure how easy this is so probably not that important
• [x] Extra reading: could add Gelman Hill to list of additional resources (chapter 9)
• [x] Extra reading II: David Freedman stuff is fun. e.g. 'statistical models and shoe leather'

[mkiang]

@MJAlexander thanks for looking. Modified your comment into a task list so it'll be easier for me to go through.

[mkiang]

Also for future reference (to myself, mostly), MathJax will allow you to do $\def\indep{\perp \! \! \perp}$ and then refer to \indep for conditional independence symbol. I couldn't get any methods that required loading packages to work.

[mkiang]

See lecture-1-update branch. @writezhe do you have examples of counterfactuals you want to include?

[mkiang]

@MJAlexander do you have a specific Sehkon paper you like? Want to add more non-health stuff.

[MJAlexander]

• [ ] My favorite is this one http://sekhon.berkeley.edu/papers/SekhonTitiunik.pdf but may not be appropriate. I'll look for another example

[mkiang]

Almost all comments and suggestions have been incorporated. See b0cb7b875c2b9ad18717dde695e7c1ca340d98e0. Still need example of good CI.

[mkiang]

Just need good example of CI now. Transitions are in. Notes are in. Could give it now if we had a CI example. Will think on it, but maybe give up on Mexico-centric example. See ea430b8fd2da10cb076c18972e56cc76b9f058a1.

[mkiang]

• [ ] Also need to add difficulty of appropriate counterfactuals. Fastest tricycle vs slowest Ferrari.

[mkiang]

Finished with L1