Timeline for How can I improve my ARIMA forecast in R?
Current License: CC BY-SA 4.0
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 15, 2019 at 8:04 | comment | added | IrishStat | If you are satisfied with my answer then please accept it to close the question. | |
| Nov 15, 2019 at 1:16 | comment | added | Santi | Thank you @IrishStat Nick Cox!! | |
| Nov 12, 2019 at 19:03 | comment | added | Nick Cox | I like the UOC framework best from a distance, but my perspective on that is like admiring Leonardo. I like the results, but I am not a painter myself. | |
| Nov 12, 2019 at 18:55 | comment | added | IrishStat | i agree with you ...and I am not suggesting anything near an OXFORD / Hendry saturation approach | |
| Nov 12, 2019 at 18:44 | comment | added | Nick Cox | Good short question but a short comment can't do it justice. I would have more faith in this procedure if it were what just about every time series expert does (manifestly, I am not one); and my impression is that it isn't. In regression models there are people who fit an indicator as an extra predictor for each marked outlier, and ad hoc is my polite name for that | |
| Nov 12, 2019 at 18:39 | comment | added | IrishStat | always good advice ...to get user suggested structure ... Do you really mean to say that discovering the need for 5 pulses and 1 level shift is dicey with 565 observations ? | |
| Nov 12, 2019 at 17:22 | comment | added | Nick Cox | If there is a monthly pattern we need to know about it to give advice. I know you prefer to automate analyses, although personally I would be sceptical about identifying numerous shifts and pulses in a short series without independent information. But there are other styles, including basing models on what is known about the generating process. As the OP wants advice, telling us what they know is likely to encourage good advice. | |
| Nov 12, 2019 at 17:17 | comment | added | IrishStat | with 20 months, I am usually reticent/reluctant to declare/test for monthly effects | |
| Nov 12, 2019 at 14:20 | comment | added | Nick Cox | My own guess is that substantive information is needed for a better, or even a good, model. For example, a very wild guess: does participation involve monthly payments or sign-ups, or otherwise why does there appear to be a monthly cycle (pun not intended)? | |
| Nov 12, 2019 at 13:22 | comment | added | Nick Cox | The term cumulant was suggested by Harold Hotelling, a US citizen. See en.wikipedia.org/wiki/Cumulant Hence I don't think this has anything to do with American English and British English. It's all to do with avoiding a term which does not have the meaning you intend in favour of a more common word which does. | |
| Nov 12, 2019 at 13:19 | history | edited | IrishStat | CC BY-SA 4.0 |
added 95 characters in body
|
| Nov 12, 2019 at 13:15 | comment | added | IrishStat | tu ... as usual we are two countries separated by a common language. | |
| Nov 12, 2019 at 12:59 | comment | added | Nick Cox | Edited all mentions of "cumulant" to "cumulative", as "cumulant" has a specific statistical meaning not applicable here (cumulants are related to moments). | |
| Nov 12, 2019 at 12:58 | history | edited | Nick Cox | CC BY-SA 4.0 |
deleted 49 characters in body
|
| Nov 12, 2019 at 12:27 | history | edited | IrishStat | CC BY-SA 4.0 |
added 130 characters in body
|
| Nov 12, 2019 at 11:34 | history | edited | IrishStat | CC BY-SA 4.0 |
added 75 characters in body
|
| Nov 12, 2019 at 11:25 | history | edited | IrishStat | CC BY-SA 4.0 |
added 75 characters in body
|
| Nov 12, 2019 at 10:15 | history | edited | IrishStat | CC BY-SA 4.0 |
added 281 characters in body
|
| Nov 12, 2019 at 3:02 | history | answered | IrishStat | CC BY-SA 4.0 |