Timeline for VAR or VECM for a mix of stationary and nonstationary variables?
Current License: CC BY-SA 4.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 27, 2023 at 12:11 | history | edited | Richard Hardy | CC BY-SA 4.0 |
added 56 characters in body
|
| Dec 12, 2022 at 16:12 | comment | added | Anthony | Ah that make sense. Thanks! | |
| Dec 12, 2022 at 15:38 | comment | added | Richard Hardy | @Anthony, the thing is, you specify these for the whole set of variables, not for subsets. So case B is not covered. | |
| Dec 12, 2022 at 15:29 | comment | added | Anthony |
Thanks, that make sense. I was just curious about the option to specify the number of cointegrating relationships, which seems to be a feature of R packages as well (rdocumentation.org/packages/tsDyn/versions/11.0.2/topics/VECM). I'll just have to do a little more reading.
|
|
| Dec 12, 2022 at 15:19 | comment | added | Richard Hardy | @Anthony, I do not use statsmodels, so I cannot tell exactly. I think this has to be handled manually. Basic cointegration routines typically assume that either all variables that are supplied are cointegrated or none are. Cases where some are cointegrated and some are not are more nuanced than that. Case B can be estimated as four separate univariate models (multiple linear regressions). | |
| Dec 12, 2022 at 15:11 | comment | added | Anthony |
Thank you for this comprehensive answer. Question re case (B): is this implemented in software by simply specifying the cointegration rank of a VECM (e.g. coint_rank in statsmodels, statsmodels.org/dev/generated/…) or does something more sophisticated have to be done?
|
|
| Nov 10, 2017 at 18:06 | history | edited | Richard Hardy | CC BY-SA 3.0 |
deleted 159 characters in body
|
| Jul 22, 2016 at 15:25 | history | edited | Richard Hardy | CC BY-SA 3.0 |
added 208 characters in body
|
| May 1, 2015 at 18:38 | history | answered | Richard Hardy | CC BY-SA 3.0 |