Annie's Chpt 15 Notes

[anniecollins]

• Line 230: briefly explain or reword what “condition on it” means here
• Two common paradoxes section: might be useful to include a bit on how each of these two situations could be identified in unfamiliar data. You mention at the beginning that DAGs can be used, so it would be good to follow up on that point, as well as emphasizing the need to graph by different groupings during EDA or understanding the sources and context for your data (for Berkson’s paradox).
• Diff in diff Simulated example: what are your intended units for serve speed here, and are these speeds/differences realistic for this scenario? I find “we find that we estimate the effect of the new tennis racket to be 5.06” a bit hard to understand for what is intended to be an illustrative example
• Section 15.4.2: Assumptions: Since the original assumptions were listed as questions, it would be good to restate the actual “parallel trends” assumption here as a statement
  • You don’t really provide any conclusion to the Raptors vs. Warriors example. What was the relationship before the stadium move versus after? Do you have evidence for or against the idea that this relationship would have remained the same if not for the move? The problem gets set up but there isn’t really any specific causal relationship proposed, nor any argument for or against this causal relationship
  • It seems a bit contradictory to, at the beginning of this section, refer to the three assumptions introduced earlier and then introduce four different assumptions/threats to validity (albeit with overlap) at the end. I would add in something that either connects these to or distinguishes them from the three assumptions/questions asked in the tennis example, so readers more clearly know what the main takeaways of the chapter are (i.e. which assumptions are the key assumptions, is there a difference between assumptions and threats to validity)
• Line 600: “As we discussed in Chapter 10, there is a trade-off between the length of the analysis that we run.” — trade-off between length of analysis and what?
• Line 1075: where is the number 371 coming from?
• Propensity score matching: Add interpretation of free shipping coefficient and relate it back to simulated data. Also include a summary of why we might use propensity score matching in the context of causal inference.
• Line 1295: what would be the forcing variable in the basketball example? I think this needs to be explained better and connected to regression discontinuity more clearly
• Exercises, Q31: What is this question supposed to be lol