Timeline for Regression hypothesis testing via out-of-sample testing
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 22, 2021 at 14:06 | answer | added | markowitz | timeline score: 1 | |
| May 21, 2021 at 18:49 | comment | added | Dave | @whuber I can see how PRESS could be used for this, and I know it has a shortcut for calculating it rather than going through the full leave-one-out-cross-validation, but it does not appear to have much relation to testing my hypothesis. | |
| May 21, 2021 at 18:47 | history | edited | Dave | CC BY-SA 4.0 |
added 72 characters in body
|
| Mar 3, 2020 at 17:25 | comment | added | whuber♦ | This sounds remarkably like PRESS, which has been around for a couple of generations. | |
| Mar 3, 2020 at 16:47 | comment | added | Dave | @jbowman "Forecasting" a time series is a little different, because the pattern could change. Maybe for the first 100 observations, AR(1) is the right model, but when you do some other model and start forecasting at $t=101, 102,\dots$, you get better performance because ARMA(1,1) is the true model for $t=101$ and beyond. Applying this to time series interpolation would be very similar to what I mean, though. In any event, that is an interesting presentation. | |
| Mar 3, 2020 at 15:42 | comment | added | jbowman | Would something like statweb.stanford.edu/~ckirby/ted/conference/… be relevant? It doesn't fully address the specific situation you are in, but insofar as "predictive accuracy" and "testing out-of-sample" are related, it may be helpful. | |
| Mar 3, 2020 at 15:34 | history | asked | Dave | CC BY-SA 4.0 |