Timeline for Questions about specifying linear mixed models in R for repeated measures data with additional nesting structure
Current License: CC BY-SA 3.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 15, 2012 at 12:44 | history | edited | Macro | CC BY-SA 3.0 |
improved title.
|
| Aug 14, 2012 at 20:53 | vote | accept | Cel | ||
| Aug 14, 2012 at 19:08 | answer | added | Macro | timeline score: 18 | |
| Aug 14, 2012 at 14:53 | history | tweeted | twitter.com/#!/StackStats/status/235388557596962817 | ||
| Aug 14, 2012 at 14:12 | comment | added | Cel | @Macro great to know, i'll include only the interactions that seem plausible then. do you have suggestions regarding the other issues? if you do, maybe put it as an answer so i could accept it. | |
| Aug 14, 2012 at 14:05 | comment | added | Macro | Hi @Cel, it looks like you've got ALL interactions in the model, including the 5-way, 4-way and 3-way interactions. I'm not sure about this case, but that will typically wildly overfit the data, which will make your results less generalizable. Backward selection (if you must use it) does not need to start with a completely saturated model - it should start with the largest model you find plausible. Can you reduce that at all? | |
| Aug 14, 2012 at 8:24 | comment | added | Cel | if i cannot find advice here i really dont know who i could ask? maybe you know of any dedicated mixed models forums or even an expert willing to consult for a little money? | |
| Aug 14, 2012 at 8:21 | history | edited | Cel | CC BY-SA 3.0 |
added 5 characters in body
|
| Aug 13, 2012 at 17:23 | history | edited | user88 | CC BY-SA 3.0 |
edited title
|
| Aug 13, 2012 at 16:26 | history | asked | Cel | CC BY-SA 3.0 |