Timeline for Testing whether there is an increase between two regression slopes in a time series
Current License: CC BY-SA 3.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 1, 2016 at 12:24 | vote | accept | Geoff | ||
| Jul 1, 2016 at 12:24 | answer | added | Geoff | timeline score: 6 | |
| Feb 14, 2013 at 13:12 | history | edited | user88 | CC BY-SA 3.0 |
edited title
|
| Feb 13, 2013 at 21:55 | answer | added | IrishStat | timeline score: -2 | |
| Feb 13, 2013 at 21:08 | comment | added | whuber♦ | "Simply" means what I described as your initial step: using two separate linear regressions, which you will conduct using ordinary least squares. (Actually, to provide a point of departure, consider conducting an initial linear regression of the entire dataset. It might perform poorly, but it will help you evaluate the extent to which more complicated models improve on it. The two-slope model is "nested" within this one, allowing the significance of the second slope to be tested--and that amounts to testing whether the two slopes differ.) | |
| Feb 13, 2013 at 21:03 | comment | added | Geoff |
you're probably in good shape. let's assume I am (I will check that tomorrow) then to start simply means I start where? I don't mind working through something, as long as I have name to start with. I appreciate your help hy the way.
|
|
| Feb 13, 2013 at 20:58 | comment | added | whuber♦ | Yes, it can be important--but isn't necessarily so. It's always a good idea to start simply and work upwards from there. In your case you ought initially to perform two separate regressions and test the residuals in each for serial correlation (your software might automatically perform a Durbin-Watson test to check that). If there's no strong evidence of serial correlation, you're probably in good shape. If the mean squared residuals are of comparable size in both regressions, then combine the data into one using a dummy variable. | |
| Feb 13, 2013 at 20:53 | comment | added | Geoff | Thanks for the info. I did find these other pages, but couldn't find one with time series, I wasn't sure if that could be important. | |
| Feb 13, 2013 at 20:52 | review | First posts | |||
| Feb 13, 2013 at 20:58 | |||||
| Feb 13, 2013 at 20:51 | comment | added | whuber♦ | Some solutions have already appeared here, Geoff, because one approach is to treat the two parts of the data as separate and regress them separately. stats.stackexchange.com/questions/33013 and other threads explain how to do that. There are additional methods, such as including a binary dummy variable to indicate the post-1980 data. I suspect these, too, have been described here: you might find solutions under the change-point tag. Finally, many time-series questions address similar issues. | |
| Feb 13, 2013 at 20:47 | comment | added | Geoff | No. Basically I'm a humble high school maths teacher and I had this idea to get such data from around the world (one station per pupil). A geography teacher suggested the date as a guide without having seen any of the data. Well, none of my data, maybe he has seen other data. | |
| Feb 13, 2013 at 20:43 | comment | added | whuber♦ | But did the colleague have any information about the sea level data? For instance, if they simply had glanced at them at one point and suggested using 1980, that alone should change the p-value of any test by a lot! | |
| Feb 13, 2013 at 20:41 | comment | added | Geoff | It comes from a non-statistical, data independent, method (a colleague just suggested it). | |
| Feb 13, 2013 at 20:38 | comment | added | whuber♦ | What was the basis for breaking the regressions at 1980? The test depends on whether that point was determined somehow from the data--even if a non-statistical method was used--or represents information independent of the data. | |
| Feb 13, 2013 at 20:37 | history | edited | whuber♦ |
edited tags
|
|
| Feb 13, 2013 at 20:35 | history | asked | Geoff | CC BY-SA 3.0 |