Timeline for Do all interactions terms need their individual terms in regression model?
Current License: CC BY-SA 3.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 4, 2012 at 19:07 | history | edited | Tristan | CC BY-SA 3.0 | deleted 1 characters in body |
| May 4, 2012 at 16:40 | comment | added | Macro | (+1) This is all true, but the original poster seemed to be describing a situation where the authors were trying to do model selection, and some of their candidate models were ones that did not include interactions - so their motivation was guided by AIC rather than by something substantive (which is always a dangerous thing to do, but apparently they've done it). When you're guided by something substantive, then the model structure should be dictated by that. But, when you're guided by a statistical criteria, leaving out main effects can have bad properties, as I've indicated in my answer. | |
| May 4, 2012 at 3:51 | history | answered | Tristan | CC BY-SA 3.0 |