NTs Are Unable to Grasp The Simple Concepts of Probability.

Post Reply New Topic

What does subjective probability mean in a non-Bayesian context?

Subjective probability is basically the degree you personally assign to something occurring, usually based off incomplete or unknown information about an event. For example, I personally believe the probability of me eating a chicken sandwich tomorrow is 60% (I don't know for sure what I'll have for lunch tomorrow, and I haven't recorded what I have eaten for lunch over time, so I can't compute all of it). The more information you know about something, the more accurate you can make predictions about it (usually by basing your subjective probability based on statistics and averages). Alternatively (and perhaps more clearly), if I asked you, "What is the probability that the U.S. and Russia will suspend diplomatic relations in 2009?", your answer would be your subjective probability.

So isn't that Bayesian no matter what, since the probability is referring to a degree of belief?

So looks like "NT" is the new "ni**er," huh? What a great way to put others in their place. What a wonderful little slur.

This is like someone saying, "Black people are obsessed with fornication."

Damn all of you for your pride and damn you for your hypocracy!

No, the Bayesians would say the prior probability is essentially the correct one. For example, which would you rank as more probable: "The U.S. and Russia will suspend diplomatic relations in 2009" or "The U.S. and Russia will suspend diplomatic relations in 2009 because of a Russian invasion of Poland". You'd be tempted to say the later because it sounds more plausible; however, it is in fact less probable. The more conditions you add to a statement may increase its plausibility while by reducing its probability (because the sample of results for the U.S. and Russia suspending diplomatic relations, which would include for the possibility various other reasons is more larger than just the one "The U.S. and Russia suspend diplomatic relations because Russia invades Poland"). Most Bayesian reasoning is not very intuitive.

I've heard similar examples before, but I've never understood how anyone could feel intuitively that the second statement sounds more likely. Evidently a lot of people must feel this way, but to me it seems intuitively obvious that the first is more likely as the set of cases it describes is a superset of the set of cases the other describes.

Anyway, what it sounds like you're talking about to me is "objective Bayesianism". That is, the idea that there is an objectively correct way to distribute prior probability.

Most generally, Bayesianism refers to the idea that probability describes degrees of belief, but there are plenty of so-called "subjective Bayesians" who don't believe there's any rational means to distribute prior probabilities (as long as they aren't 1 or 0), and bank on the fact that with enough data the priors effectively end up getting "washed out".

As it happens, I am an objective Bayesian in the sense that I believe some priors at least make more sense than others, and that there should be a negative relationship between the amount of information a proposition asserts (by whatever information measure) and the prior probability it is assigned. But what I was getting at was that Bayesianism is more general than objective Bayesianism, and any interpretation of probability as a degree of belief is Bayesian.

True, I was describing objective Bayesian (which I view has having more use than subjective Bayesian probability because well, it is purely subjective so is only useful in certain situations). I'd agree that general Bayesianism even without the 2 schools is an interpretation of the probability as a degree of something occurring or being true.