From appendix B of Regression and Other Stories:
1. Think about variation and replication
2. Forget about statistical significance
3. Graph the relevant and not the irrelevant
4. Interpret regression coefficients as comparisons
5. Understand statistical methods using fake-data simulation
6. Fit many models
7. Set up a computational workflow
8. Use transformations
9. Do causal inference in a targeted way, not as a byproduct of a large regression
10. Learn methods through live examples
Very nice list.
I especially applaud #2
I hope #6 is meant as fitting many models to the same data, which is what I’d advocate
Paul:
Yes, fitting many models to the same data was what I was thinking of. Fitting models to new data is good too!
Once you agree with #6, #2 loses interpretable meaning anyway!
If there is one lesson I try to teach in my graduate applied economics class, it is “Interpret regression coefficients as comparisons.”
It does so many useful things for students, but I think one of the under-appreciated benefits is that it can help to de-mystify statistical analysis more generally. Statistical analyses aren’t incantations or spells that can magically reveal hidden things about the world – they are just ways to quantify comparisons of outcome variables (one group to another, or low and high values of X).
If there is one lesson I try to teach in my graduate applied economics class, it is “Interpret regression coefficients as comparisons.”
It does so many useful things for students, but I think one of the under-appreciated benefits is that it can help to de-mystify statistical analysis more generally. Statistical analyses aren’t incantations or spells that can magically reveal hidden things about the world – they are just ways to quantify comparisons of outcome variables (one group to another, or low and high values of X).
I’m now imagining your counterpart at Hogwarts struggling to “re-mystify” incantations for his students!
Expecto pearsonum.
Very late to this party but I want to post my two bits. I don’t even know how this works – can I address Andrew? I encountered Andrew many many years ago while I was a graduate student at UC, Berkeley. I have very little recollection except vaguely remembering someone telling me – that’s Andrew Gelman, he’s a Bayesian you know, with raised eyebrows. I was taking Leo’s Breiman’s applied statistics class at that time.
Flash forward to now and I’m working with data which has complex variability structure and I am cast to the past to Leo Breiman’s class when he often scratched his head, pacing up and down the class repeating – Where’s the randomness, where’s the randomness coming from. Back in the present, I am asking all my colleagues the same thing – where’s the randomness coming from, that’s what we need to know. And through this quest, I re-encounter Andrew Gelman and flipping through the workmanship I see – tip #1 Think about variation and replication.
I am very excited now to learn from a Bayesian. Better late than never is correct. Love everything I discover, especially – Yes, I really really really like fake-data simulation, and I can’t stop talking about it.(can’t link it) – and other stories.
Cheers