Timeline for How does variance-inflation factor creep into a chunk test?
Current License: CC BY-SA 4.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 14, 2023 at 16:16 | history | edited | Dave | CC BY-SA 4.0 |
added 218 characters in body
|
| S Jan 25, 2023 at 17:37 | history | bounty ended | Dave | ||
| S Jan 25, 2023 at 17:37 | history | notice removed | Dave | ||
| Jan 22, 2023 at 20:36 | answer | added | EdM | timeline score: 2 | |
| Jan 18, 2023 at 20:17 | comment | added | whuber♦ | @BigBendRegion Great comments. Much of the interesting math to which you refer was done in 1980 by Belsley, Kuh, & Welsch, Regression diagnostics. They focus on finding groups of variables that might contribute individually to multicollinearity, yet be relatively uncorrelated between groups. | |
| S Jan 18, 2023 at 16:23 | history | bounty started | Dave | ||
| S Jan 18, 2023 at 16:23 | history | notice added | Dave | Draw attention | |
| Dec 2, 2022 at 12:55 | comment | added | BigBendRegion | A "chunk" can be highly significant while variables within the chunk are insignificant, due to multicollinearity of variables within the chunk. So variance inflation does not seem to be an issue for the chunk test, despite collinearity within the chunk. As an example, there is perfect multicollinearity between variables in the full dummy representation of a factor, but the partial F test is not affected. On the other hand, if there is multicollinearity between chunks, there will definitely be a variance inflation issue. Seems like interesting math involving a cross- correlation matrix. | |
| Dec 2, 2022 at 5:52 | history | asked | Dave | CC BY-SA 4.0 |