New answers tagged

1

The “typical” Bayesian way to do this (at least, it seems typical to someone with my background – I’m sure others would greatly disagree) would go like this: There is some unknown function $f(\mathbf{x})$, where $\mathbf{x}$ is a four-element vector. You have (possibly noisy) observations of $f(\mathbf{x}_i) = y_i$ for $i = 1, \dots, N$. You want to find an ...


1

Following on from the comments, you'll still need to have some sort of model. You could have a prior for each variable and update them with your data, and that might tell you about the distributions that generated the variables, but it doesn't tell you anything about how they're linked together. Typically, a Bayesian approach would put priors over the ...


Top 50 recent answers are included