I highly recommend that everyone reading my post here should get familiar with the work of Professor Deborah Mayo. She is a genius in my opinion — her three texts written since 1996 are gems. Both are lucid but detailed and highly articulate; her writing addresses the present state of affairs vis-a-vis statistical hypothesis evaluation and scientific research. Start with her 1996 book, “Error and the Growth of Experimental Knowledge”. You will be rewarded.
The points here are bizarre at worst and at best simply out of touch with current statistical and social science practice. The criticisms of over-reliance on p-values, linear models, Diff-in-Diffs, etc are all 100% correct....if you stopped keeping up-to-date circa 2005.
Economics and political science-- two heavy consumers and producers of statistical methodology--have greatly advanced the state of art and science and practice of statistics. Just skim the writings of Andrew Gelman (Columbia) or Guido Imbens and Susan Athey (Stanford).
For posterity. This comment is mainly for people who'll stumble across this post and think statistical practice is stuck in 2000s p-hacking and GLM purgatory.
There are many interesting cases from Econ and political science. However, many experts in those fields still use traditional stat(before 2005) and some even don’t rely on stat methods. They apply other applied math tools -combinatorics/graph /PDE…which is not mentioned through Andrew Gelman. Could you point a specific state of art stat methods’ example ?
I highly recommend that everyone reading my post here should get familiar with the work of Professor Deborah Mayo. She is a genius in my opinion — her three texts written since 1996 are gems. Both are lucid but detailed and highly articulate; her writing addresses the present state of affairs vis-a-vis statistical hypothesis evaluation and scientific research. Start with her 1996 book, “Error and the Growth of Experimental Knowledge”. You will be rewarded.
The points here are bizarre at worst and at best simply out of touch with current statistical and social science practice. The criticisms of over-reliance on p-values, linear models, Diff-in-Diffs, etc are all 100% correct....if you stopped keeping up-to-date circa 2005.
Economics and political science-- two heavy consumers and producers of statistical methodology--have greatly advanced the state of art and science and practice of statistics. Just skim the writings of Andrew Gelman (Columbia) or Guido Imbens and Susan Athey (Stanford).
Weird, I've never heard of Andrew Gelman, Guido Imbens, or Susan Athey. Are they TikTok influencers?
:-)
scenario: TikTok challenge for statistician—causal inference through AGI weights —human extinction analysis
For posterity. This comment is mainly for people who'll stumble across this post and think statistical practice is stuck in 2000s p-hacking and GLM purgatory.
There are many interesting cases from Econ and political science. However, many experts in those fields still use traditional stat(before 2005) and some even don’t rely on stat methods. They apply other applied math tools -combinatorics/graph /PDE…which is not mentioned through Andrew Gelman. Could you point a specific state of art stat methods’ example ?