Timeline for Analyzing logistic regression coefficients
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 28, 2017 at 20:36 | history | edited | kjetil b halvorsen♦ | CC BY-SA 3.0 |
deleted 34 characters in body
|
| Mar 20, 2012 at 17:15 | vote | accept | shiu6rewgu | ||
| Mar 14, 2012 at 18:12 | comment | added | shiu6rewgu | Thanks all of you for your input! I really appreciate each and every one of you! A lot of my questions were answered | |
| Mar 14, 2012 at 18:11 | vote | accept | shiu6rewgu | ||
| Mar 14, 2012 at 18:11 | |||||
| Mar 14, 2012 at 18:11 | vote | accept | shiu6rewgu | ||
| Mar 14, 2012 at 18:11 | |||||
| Mar 10, 2012 at 11:38 | history | tweeted | twitter.com/#!/StackStats/status/178444673369653248 | ||
| Mar 10, 2012 at 9:10 | answer | added | conjugateprior | timeline score: 12 | |
| Mar 10, 2012 at 6:19 | answer | added | gung - Reinstate Monica | timeline score: 20 | |
| Mar 10, 2012 at 1:26 | comment | added | rolando2 | Regarding the coefficient of zero: I could see this happening as an artifact of putting all your coefficients into XL before pasting them here--something that seems consistent with the high number of decimal places we're generally seeing. Maybe one of those XL cells was set to round to integers, yielding the zero. I've had things like this happen. | |
| Mar 9, 2012 at 23:38 | comment | added | Peter Flom | Showing a list of coefficients like this, with no context at all, is likely saying "Joe has 31, isn't that a lot?" without saying 31 what. 31 cars? A lot. 31 kids? One heck of a lot! 31 dollars? Not much. | |
| Mar 9, 2012 at 21:01 | comment | added | whuber♦ | If you were to subtract 1027 from all the values of the sixth variable, your intercept would be quite close to 0. Would that make you feel better? :-) | |
| Mar 9, 2012 at 20:10 | comment | added | probabilityislogic | What is the range or standard deviation of your other variables? Is there a large difference between the standard deviation of the variable with zero estimate compared to the others? You may expect a coefficient of zero if the standard deviation is small compared to the others (numerical precision). Also intercept basically means you have variables which have large averages (away from zero). Centering your variables would give a more interpretable intercept, and won't change the betas for the other variables (iterative algorithm error aside). | |
| Mar 9, 2012 at 19:40 | history | asked | shiu6rewgu | CC BY-SA 3.0 |