"Feels cringe to recommend my own book", but actually being half way through it now and having consumed a lot of other bayesian learning material I genuinely believe it's the best starting point by miles. If you're a bayesian and you haven't read statistical rethinking then it's an absolute must.
Great talk as always. One important reason to teach frequentist thinking thoroughly is that people are inevitably going to have to read papers with (badly implemented and incorrectly interpeted) frequentist models, and as a reader of such papers I need to have fluency in the methodology to be able to determine how far the authors really are from what they want to conclude from the data. If people only know Bayes, they are not going to be able to evaluate published work or work they have to review for journals, etc.
@rmcelreath any good reference illustrating Bayesian statistics w/o frequentist language? Particularly from the pre-data perspective of the DGP as a joint probability distribution of "equivalent" variables?
Basically, it's the thing you believe before you observe anything. Priors are gonna have some parameters and you can use certain rules to update that prior based on observations, to refine your belief to better align with reality.
This presentation confused the light out of me. I cannot find any correlation between the title and the talk. Was this intended to be a joke? Probably.
Methodology, my lad, is the study of nethods in order to discover the best method for the matters at hand. Yes, I am a polyglot. It is not a top shelf synonym for methods. PS, it's not how many notes but the bird calls that indicate whether or not a cat is present.
"Feels cringe to recommend my own book", but actually being half way through it now and having consumed a lot of other bayesian learning material I genuinely believe it's the best starting point by miles. If you're a bayesian and you haven't read statistical rethinking then it's an absolute must.
I started with kruschke and I don't think it was a waste of time but stat. Rethinking was way superior in many ways.
Great talk as always. One important reason to teach frequentist thinking thoroughly is that people are inevitably going to have to read papers with (badly implemented and incorrectly interpeted) frequentist models, and as a reader of such papers I need to have fluency in the methodology to be able to determine how far the authors really are from what they want to conclude from the data. If people only know Bayes, they are not going to be able to evaluate published work or work they have to review for journals, etc.
@rmcelreath any good reference illustrating Bayesian statistics w/o frequentist language? Particularly from the pre-data perspective of the DGP as a joint probability distribution of "equivalent" variables?
👍
12:23
What is a prior?
--
Lookup "prior probability".
Basically, it's the thing you believe before you observe anything.
Priors are gonna have some parameters and you can use certain rules to update that prior based on observations, to refine your belief to better align with reality.
This presentation confused the light out of me. I cannot find any correlation between the title and the talk. Was this intended to be a joke? Probably.
Methodology, my lad, is the study of nethods in order to discover the best method for the matters at hand. Yes, I am a polyglot. It is not a top shelf synonym for methods. PS, it's not how many notes but the bird calls that indicate whether or not a cat is present.