Timeline for Why Normality assumption in linear regression
Current License: CC BY-SA 4.0
23 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 7, 2019 at 14:30 | comment | added | JiK | @AdamO Models are used in real life too. | |
| Mar 4, 2019 at 12:00 | history | tweeted | twitter.com/StackStats/status/1102539291811549185 | ||
| Mar 4, 2019 at 11:13 | history | reopened |
Sextus Empiricus jbowman Peter Flom regression Users with the regression badge or a synonym can single-handedly close regression questions as duplicates and reopen them as needed. |
||
| Mar 4, 2019 at 0:15 | review | Reopen votes | |||
| Mar 4, 2019 at 11:13 | |||||
| Mar 4, 2019 at 0:01 | comment | added | Sextus Empiricus | @kjetilbhalvorsen I don't see how this is a duplicate. This question is about using different assumptions than the normality assumption for the error term. This is not the case in the duplicate question which is about "(the marginal) X and Y are non-normal but the error term is". | |
| Mar 3, 2019 at 22:22 | history | closed |
kjetil b halvorsen♦ AdamO COOLSerdash Robert Long Taylor |
Duplicate of Normality assumption in linear regression | |
| Mar 2, 2019 at 21:54 | comment | added | AdamO | @jik it's called real life. You collect data with a scientific question in mind and discern whether a prespecified analysis is capable of answering that question. Very different from textbooks. | |
| Mar 2, 2019 at 18:12 | comment | added | JiK | @AdamO I'd love to read more about doing statistical inference without any assumptions. | |
| Mar 2, 2019 at 15:03 | answer | added | Neil G | timeline score: -2 | |
| Mar 2, 2019 at 12:46 | answer | added | Sextus Empiricus | timeline score: 2 | |
| Mar 2, 2019 at 2:41 | comment | added | smci | @kjetilbhalvorsen: Both titles are similar, but the question bodies ask "Why we assume normal distribution of error terms?" vs "Can we construct a scenario where residuals are normally distributed but X, Y are not?" vs "What if residuals are normally distributed but Y is not?", which itself is a further near-duplicate. Could you users with enough rep here please start fixing titles and aggressively closing duplicates? | |
| Mar 2, 2019 at 0:05 | answer | added | David | timeline score: 2 | |
| Mar 1, 2019 at 22:39 | comment | added | AdamO | @JiK if I were choosing assumptions, I would choose none at all. It turns out OLS is a minimax estimator that minimizes squared error loss and that is very useful. The only reason a "normal" error is useful is that you can calculate an exact F-test for the significance of model coefficients. In decent sample sizes, even that doesn't matter. OLS is quite robust to non-normal errors by the CLT. Even Gauss noted this almost 200 years ago when he derived the OLS estimator, but this fact seems to be lost to history in the overly simplistic way that we now teach regression modeling. | |
| Mar 1, 2019 at 21:45 | comment | added | JiK | @AdamO You can choose assumptions for your model when you're doing statistical inference, so I don't think that means there is no statistics. | |
| Mar 1, 2019 at 17:26 | answer | added | Martin L | timeline score: 10 | |
| Mar 1, 2019 at 15:40 | comment | added | AdamO | @JiK if I could choose distributions, there'd be no need for statistics at all. The whole world would be probability. | |
| Mar 1, 2019 at 14:34 | comment | added | JiK | @AdamO I don't understand; you just outlined the reasons why we choose it. | |
| Mar 1, 2019 at 12:25 | review | Close votes | |||
| Mar 1, 2019 at 12:37 | |||||
| Mar 1, 2019 at 6:05 | history | became hot network question | |||
| Mar 1, 2019 at 4:32 | comment | added | Nat | Because the math works out easily enough that people could use it before modern computers. | |
| Mar 1, 2019 at 4:25 | comment | added | AdamO | We don't choose the normal assumption. It just happens to be the case that when the error is normal, the model coefficients exactly follow a normal distribution and an exact F-test can be used to test hypotheses about them. | |
| Mar 1, 2019 at 4:14 | answer | added | Glen_b | timeline score: 32 | |
| Mar 1, 2019 at 3:54 | history | asked | Master Shi | CC BY-SA 4.0 |