How do you conduct these interviews? I would've expected every question provided at the start considering the long answers - and if someone was sending me one question at a time I'd wonder if they were leading me somewhere.
But from the back and forth it does sound like it went one email at a time.
This is a brilliant interview/exposition. Get away from the fancy implausible econometric tricks. Stick to problems where the data set you are using is robust enough to do the heavy lifted. Pay attention to the inherent quality and reliability of your data. Use common sense. Bravo!
Glad you brought up Breiman's point. "The dude has blind spots"?? Indeed, but he clearly identified the cultural split in the learning community between the AI guys, who are methodologically pure, but get things all wrong, the learning machinists, who do predict stuff, and his own statistics community, who pooh on both sides.
How do you conduct these interviews? I would've expected every question provided at the start considering the long answers - and if someone was sending me one question at a time I'd wonder if they were leading me somewhere.
But from the back and forth it does sound like it went one email at a time.
This is a brilliant interview/exposition. Get away from the fancy implausible econometric tricks. Stick to problems where the data set you are using is robust enough to do the heavy lifted. Pay attention to the inherent quality and reliability of your data. Use common sense. Bravo!
Glad you brought up Breiman's point. "The dude has blind spots"?? Indeed, but he clearly identified the cultural split in the learning community between the AI guys, who are methodologically pure, but get things all wrong, the learning machinists, who do predict stuff, and his own statistics community, who pooh on both sides.
Re: "collapsing the wave function", I like Stephen Senn's example here https://errorstatistics.com/2016/08/02/s-senn-painful-dichotomies-guest-post/, which is less multifarious than the Stent paper Gelman mentions but makes a single point very clearly.
Great interview!