Timeline for A more definitive discussion of variable selection
Current License: CC BY-SA 3.0
26 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 13, 2017 at 12:44 | history | edited | CommunityBot |
replaced http://stats.stackexchange.com/ with https://stats.stackexchange.com/
|
|
| S Jul 23, 2016 at 20:57 | history | bounty ended | gung - Reinstate Monica | ||
| S Jul 23, 2016 at 20:57 | history | notice removed | gung - Reinstate Monica | ||
| Jul 21, 2016 at 19:19 | answer | added | David C. Norris | timeline score: 11 | |
| Jul 16, 2016 at 21:12 | vote | accept | sharper_image | ||
| S Jul 16, 2016 at 20:31 | history | bounty started | gung - Reinstate Monica | ||
| S Jul 16, 2016 at 20:31 | history | notice added | gung - Reinstate Monica | Reward existing answer | |
| Jul 15, 2016 at 11:23 | history | tweeted | twitter.com/StackStats/status/753912933772591104 | ||
| Jul 15, 2016 at 10:17 | comment | added | amoeba | @Scortchi I am surprised though to see any recommendation based on p-values in a Gelman's text; he must have written hundreds of blog posts about how p-values are misleading and unhelpful. | |
| Jul 15, 2016 at 9:48 | comment | added | Scortchi♦ | Re Gelman & Hill's 4(b): Harrell, RMS 1st edn, p60: "A final problem with variable selection is illustrated by comparing this approach [sc. stepwise] with the sensible way many economists develop regression models. Economists frequently use the strategy of deleting only those variables that are insignificant & whose regression coefficients have a nonsensible direction". And I seem to recall that Steyerberg was going to write a paper on it. [Will try to find the ref. when I get chance.] | |
| Jul 14, 2016 at 23:10 | comment | added | amoeba | See also an excellent and a very popular answer by @gung here stats.stackexchange.com/questions/20836 about the pitfalls of stepwise approaches. | |
| Jul 14, 2016 at 20:25 | history | edited | sharper_image | CC BY-SA 3.0 |
removed sentence that was no longer relevant in background
|
| Jul 14, 2016 at 20:20 | history | edited | sharper_image | CC BY-SA 3.0 |
added MARS hyperlink in response to comment
|
| Jul 14, 2016 at 18:53 | comment | added | gung - Reinstate Monica | Harrell, Flom, & Kolassa; it's settled then. | |
| Jul 14, 2016 at 18:04 | comment | added | Stephan Kolassa | Not only Harrell and Flom dislike stepwise regression. I do so, too. Very nice first question, BTW! | |
| Jul 14, 2016 at 17:27 | answer | added | AdamO | timeline score: 51 | |
| Jul 14, 2016 at 17:06 | comment | added | user78229 | +1 for a very thoughtful question. One thing your review demonstrates is that there is no standard recipe or cookbook for variable selection. Everybody has a different approach and set of "best" practices. Then, there is the very important distinction between a theoretical or population model vs an empirically driven one--these different frameworks rarely align and it's easy to conflate them. Your concern about "missing" something is not unfounded but the reality is that a conclusive, unequivocal answer simply does not exist. The approaches you've reviewed are extensively documented, pick one | |
| Jul 14, 2016 at 16:59 | history | edited | sharper_image | CC BY-SA 3.0 |
added 8 characters in body
|
| Jul 14, 2016 at 16:55 | comment | added | sharper_image | That is a good point; given that I learned (or at least recall learning) to do it manually in that fashion, it just seemed reminiscent of doing forward stepwise regression manually or best subset regression based on AIC/BIC. This is perhaps due to my own ignorance though which I would love to be corrected on! Also: appreciate your moving the links in! | |
| Jul 14, 2016 at 16:53 | history | edited | gung - Reinstate Monica | CC BY-SA 3.0 |
added tags; formatted; light editing; added links
|
| Jul 14, 2016 at 16:49 | comment | added | gung - Reinstate Monica | Whether or not something can be automated (at present--prior to the advent of strong AI) strikes me as a red herring. | |
| Jul 14, 2016 at 16:45 | history | edited | gung - Reinstate Monica | CC BY-SA 3.0 |
added tags; formatted; light editing
|
| Jul 14, 2016 at 16:35 | history | edited | sharper_image | CC BY-SA 3.0 |
added 15 characters in body
|
| Jul 14, 2016 at 16:31 | review | First posts | |||
| Jul 14, 2016 at 16:44 | |||||
| Jul 14, 2016 at 16:30 | comment | added | sharper_image | References 1. stats.stackexchange.com/questions/56725/… 2. stats.stackexchange.com/questions/221907/… 3. stats.stackexchange.com/questions/18214/… 4. lexjansen.com/pnwsug/2008/DavidCassell-StoppingStepwise.pdf 5. andrewgelman.com/2014/06/02/hate-stepwise-regression 6. stats.stackexchange.com/questions/138860/… | |
| Jul 14, 2016 at 16:30 | history | asked | sharper_image | CC BY-SA 3.0 |