Timeline for Omitted variable bias in linear regression
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| S Jul 11, 2017 at 13:34 | history | suggested | Steve | CC BY-SA 3.0 |
fixed math notation
|
| Jul 11, 2017 at 13:06 | review | Suggested edits | |||
| S Jul 11, 2017 at 13:34 | |||||
| Feb 14, 2013 at 23:32 | vote | accept | Josu Momediano | ||
| Feb 14, 2013 at 22:24 | answer | added | gung - Reinstate Monica | timeline score: 13 | |
| Feb 14, 2013 at 22:16 | comment | added | Michael Bishop | You are correct to be concerned. A lot of inference is based on the assumption that we have the true model. I've been running regressions a long time and I've never had the true model. For my purposes it rarely makes sense to even think that one true model exists. Instead, ask yourself what the goals of your modelling are (prediction in sample, prediction out of sample, estimating the average causal effect of x3, data summary, etc.) because your goals will indicate which modelling strategies are best. | |
| Feb 14, 2013 at 21:37 | comment | added | gung - Reinstate Monica | Also, my answer here: identifying-the-population-and-samples-in-a-study, is tangentially related, & may be of interest. | |
| Feb 14, 2013 at 21:21 | comment | added | Josu Momediano | But whatever the method you use (expertise included), you begin from 0, and you are 100% having the problem I talk about... It's like omitted variable bias is there every time | |
| Feb 14, 2013 at 21:12 | history | edited | user88 | CC BY-SA 3.0 |
edited title
|
| Feb 14, 2013 at 21:10 | comment | added | gung - Reinstate Monica | As a side note, stepwise selection methods (such as forward stepwise) are very unlikely to pick out the model that you ought to be using. If this doesn't make sense, you may want to read my answer here: algorithms-for-automatic-model-selection. | |
| Feb 14, 2013 at 21:00 | history | edited | gung - Reinstate Monica | CC BY-SA 3.0 |
added tags; formatted; edited for smoother English
|
| Feb 14, 2013 at 20:44 | comment | added | Josu Momediano | Yeah, I meant that. You have the sample/observations, and then the conditions (Gauss-Markov), which guarantee the estimators to be the best unbiassed ones etc | |
| Feb 14, 2013 at 20:42 | history | edited | Josu Momediano | CC BY-SA 3.0 |
edited title
|
| Feb 14, 2013 at 20:28 | comment | added | Andy W |
In the part samples from (Y,X1,....Xn) and then a bunch of conditions by which the OLS estimations behave quite well. is that really what you meant or did some of your sentence get cut off. Also you have a misspelling in the title of the question.
|
|
| Feb 14, 2013 at 20:23 | review | First posts | |||
| Feb 14, 2013 at 21:00 | |||||
| Feb 14, 2013 at 20:05 | history | asked | Josu Momediano | CC BY-SA 3.0 |