This guy’s video is a fruit salad. It is painful to watch so many concepts mixed and misrepresented. To anyone watching, please take this video as an uninformed opinion and read Jaynes “Probability Theory the logic of Science.” Or Gregory “Bayesian Logical Data Analysis for the Physical Sciences”
He might be knowledgable or awesome to those people with sufficient prior knowledge but tbh I don't think mathematics could be discussed in such a general and vague way without clear fomular on boards
+Vector Shift None. None at all. Fisher himself regarded the idea as "inherently wrong", and so did Popper. Bayesians seem to think that Frequentists are a bunch of morons who would never even consider looking at prior history or past occurrences of an event, they actually believe they are the only ones who keep past experiences in mind. There is literally nothing that gives credit to the Bayesian interpretation of probability. It relies on even more assumptions than the Frequentist view. Stick to robust statistics, it's the key.
+Vector Shift Well, as I understand it, the content of the Bayesian interpretation of probability is a recipe for modifying your an assumed likelihood of a hypothesis in light of additional information and clearly at this stage the information gained from this has had a great deal of success in practical applications. There was a claim for a long time (Cox's theorem) that Bayesian probability was none other than the single consistent generalization of Boolean logic to propositions that could take real, intermediate values between 0 (False) and 1 (True), these values taken to correspond to how much one could reasonable think the proposition to be true given prior information. Reading around, I get the impression that things are not so simple, and that perhaps Bayesian probability may just be the 'simplest' possible generalization of Boolean logic, not the only one.
+77Fortran " There was a claim for a long time (Cox's theorem) that Bayesian probability was none other than the single consistent generalization of Boolean logic to propositions that could take real, intermediate values between 0 (False) and 1 (True), these values taken to correspond to how much one could reasonable think the proposition to be true given prior information. " Isn't that just the function of ordinary probabilities?
Vector Shift I may be wrong but I get the impression that to some statisticians probabilities can only rigorously be interpreted in terms of the proportion of times something would occur in some hypothetical setup e.g. if it's about the probability of picking a white marble from an urn, imagine a colossal ensemble of such urns; over many trials, the proportion of times you selected a white ball should asymptote to the true probability. If instead probability is the measure of 'reasonable expectation given the data', then you can apply probability theory to situations of reasoning where there is no interpretation in terms of an ensemble.
This guy’s video is a fruit salad. It is painful to watch so many concepts mixed and misrepresented. To anyone watching, please take this video as an uninformed opinion and read Jaynes “Probability Theory the logic of Science.” Or Gregory “Bayesian Logical Data Analysis for the Physical Sciences”
This guy is awesome.
fantastic!
thanks
He might be knowledgable or awesome to those people with sufficient prior knowledge but tbh I don't think mathematics could be discussed in such a general and vague way without clear fomular on boards
are there any decent arguments for Bayesian "probability?
+Vector Shift None. None at all. Fisher himself regarded the idea as "inherently wrong", and so did Popper. Bayesians seem to think that Frequentists are a bunch of morons who would never even consider looking at prior history or past occurrences of an event, they actually believe they are the only ones who keep past experiences in mind. There is literally nothing that gives credit to the Bayesian interpretation of probability. It relies on even more assumptions than the Frequentist view. Stick to robust statistics, it's the key.
+Vector Shift Well, as I understand it, the content of the Bayesian interpretation of probability is a recipe for modifying your an assumed likelihood of a hypothesis in light of additional information and clearly at this stage the information gained from this has had a great deal of success in practical applications.
There was a claim for a long time (Cox's theorem) that Bayesian probability was none other than the single consistent generalization of Boolean logic to propositions that could take real, intermediate values between 0 (False) and 1 (True), these values taken to correspond to how much one could reasonable think the proposition to be true given prior information.
Reading around, I get the impression that things are not so simple, and that perhaps Bayesian probability may just be the 'simplest' possible generalization of Boolean logic, not the only one.
+77Fortran " There was a claim for a long time (Cox's theorem) that Bayesian probability was none other than the single consistent generalization of Boolean logic to propositions that could take real, intermediate values between 0 (False) and 1 (True), these values taken to correspond to how much one could reasonable think the proposition to be true given prior information. "
Isn't that just the function of ordinary probabilities?
Vector Shift I may be wrong but I get the impression that to some statisticians probabilities can only rigorously be interpreted in terms of the proportion of times something would occur in some hypothetical setup
e.g. if it's about the probability of picking a white marble from an urn, imagine a colossal ensemble of such urns; over many trials, the proportion of times you selected a white ball should asymptote to the true probability.
If instead probability is the measure of 'reasonable expectation given the data', then you can apply probability theory to situations of reasoning where there is no interpretation in terms of an ensemble.
why does it even exist?