Timeline for Why should I be Bayesian when my dataset is large?
Current License: CC BY-SA 4.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 10, 2020 at 13:09 | comment | added | kutschkem | @jnez71 Yes it was just a joke, of course there are problems where your dataset can be large (enough). But a lot of problems have such a big feature space that it's basically impossible that the dataset is "large enough". As an example, I work in automotive and they are heavily looking into simulating situations that in a "normal" dataset would almost never appear but are important nonetheless (basically any accident situation, I guess). | |
| Oct 9, 2020 at 18:08 | comment | added | jnez71 | @kutschkem whoops just realized you may have been making a joke about the OP's notion of ever having a large dataset lol | |
| Oct 9, 2020 at 18:04 | comment | added | jnez71 | Overall I think my comment is just an example of the kind of behavior pointed out in #6 of the accepted answer (by Bernhard). | |
| Oct 9, 2020 at 17:57 | comment | added | jnez71 | @kutschkem in the demo from that other rather unrelated question, yes the data set is small; it was just for visualization purposes. If the landmarks are actually all in a straight line though, the uncertainty will never reduce. And perhaps even if the deviations from a straight line are much smaller than the error variance of the angle measurements. | |
| Oct 9, 2020 at 8:12 | comment | added | kutschkem | Your dataset is never large ;-) | |
| Oct 8, 2020 at 16:41 | comment | added | jnez71 | Check out the last gif in this answer for a visualization of that Bayesian behavior. One cool thing about Bayesian reasoning is pretty much that is doesn't (necessarily) behave the way your question suggests. The remaining uncertainty in one's posterior can make clear what your data can't seem to tell you, no matter how much you get. This can be a specific combination of the variables, or some equi-dimensional irreducible variance. It tends to be very relevant for timeseries models, check out "observability analysis." | |
| Oct 8, 2020 at 16:32 | comment | added | jnez71 | "However, the prior's influence diminishes (to zero?) as the dataset grows larger." -- This is not true in general. Every new data point is not necessarily "more useful." It can keep telling you about what you already know, and not about what you don't. For example, imagine that you're trying to estimate your position by measuring your bearing (angle) to a bunch of known landmarks you can see in the distance. Your prior is a Gaussian centered where you think you are. If all of the landmarks are roughly co-linear, your posterior will never reduce uncertainty / spread along that axis. | |
| Oct 6, 2020 at 23:12 | vote | accept | kennysong | ||
| Oct 6, 2020 at 21:33 | answer | added | Tim | timeline score: 8 | |
| Oct 6, 2020 at 21:03 | answer | added | Wayne | timeline score: 8 | |
| Oct 6, 2020 at 20:29 | comment | added | Adrian | What is the goal of your analysis? Inference? Accurate predictions? Something else? | |
| Oct 6, 2020 at 19:15 | answer | added | daniel.s | timeline score: 8 | |
| Oct 6, 2020 at 15:00 | history | tweeted | twitter.com/StackStats/status/1313494253356679168 | ||
| Oct 6, 2020 at 14:41 | history | became hot network question | |||
| Oct 6, 2020 at 7:00 | answer | added | Bernhard | timeline score: 41 | |
| Oct 6, 2020 at 6:38 | history | asked | kennysong | CC BY-SA 4.0 |