Talk:Stepwise regression
 | This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: |
||||||||||
| |||||||||||
Thank god for this. Now the universe makes sense.
dumb it down?[edit]
Since a lot of people without graduate degrees in statistics use Wikipedia, I might humbly suggest that this article is supplemented with a paragraph that can explain stepwise regression in simple terms. I have seen far more complicated subjects described for novices, so I'm pretty sure someone here can do it. --Insightfullysaid — Preceding unsigned comment added by 208.90.11.207 (talk) 20:05, 20 August 2012 (UTC)
more recent developments[edit]
It'd be nice if someone with more expertise than me in the area could put something about more recent developments. I'm thinking of things like Tibshirani's lars and lasso. --Delirium 04:16, 16 October 2007 (UTC)
last point in Criticism is from 1981[edit]
A simulation study from 1981 seems to be way out of date. pointing to a problem with forward selection is highly dependent on the procedure used at the time! I hope to extend the article some at a later time Talgalili (talk) 20:35, 3 July 2009 (UTC)
Why is this called stepwise regression?[edit]
I know this as stepwise feature selection. It's not specific to regression; at least not to metric regression (real-valued dependent variable). I know regression has a broader meaning in statistics than in machine learning, but shouldn't this article be renamed "stepwise variable selection" or similar? QVVERTYVS (hm?) 20:04, 13 July 2015 (UTC)
Advertorial[edit]
Reference 5, "Stopping stepwise: Why stepwise and similar selection methods are bad, and what you should use", which is referenced twice in the article has all the signs of being "literature" paid for by SAS to promote its products. References to it should IMO be removed, and ideally replaced with something from a serious journal. — Preceding unsigned comment added by 94.109.187.250 (talk) 17:21, 16 April 2018 (UTC)
The description for bidirectional provides almost no information[edit]
The other two descriptions don't give details, but the general process is clear. For bidirectional, it is completely unclear. It says that it's a combination of the other two approaches. Forward selection starts with no variables, and backward selection starts with all of the variables. What's the combination of all and none - some? UsernamesEndedYearsAgo (talk) 00:51, 20 March 2024 (UTC)