Timeline for How is the augmented Dickey–Fuller test (ADF) table of critical values calculated?
Current License: CC BY-SA 3.0
22 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 29, 2020 at 7:01 | comment | added | Christoph Hanck | As the distribution is an asymptotic distribution, $T$ does not matter anymore in the limit. But you are of course right that finite sample distributions may differ from asymptotic ones (you could google MacKinnon and response surface regressions for some results in this regard). And indeed, as my linked post stats.stackexchange.com/questions/224084/… highlights, the distribution depends on the trend specification. | |
| Jul 28, 2020 at 18:54 | comment | added | confused | @ChristophHanck Thanks for the response! If you don't mind answering, other than what seems like T (which I believe is sample size) are there any other variables that may cause the distribution to differ from test to test? My math is not that great but just from looking at my own results, the distribution does seem to differ if you have drift or not. I was hoping to use the ADF statistic itself as a measure of relative stationarity (between two pairs of processes) and want to make sure everything is standardized. | |
| Jul 28, 2020 at 4:16 | comment | added | Christoph Hanck | @confused, unlike the t-distribution, the DF distribution is not charcterized by any degree of freedom parameter. | |
| Jul 27, 2020 at 10:26 | comment | added | confused | @ChristophHanck What is the degrees of freedom for the distribution? Is it typical n - number of coefficients? | |
| Apr 13, 2017 at 12:44 | history | edited | CommunityBot |
replaced http://stats.stackexchange.com/ with https://stats.stackexchange.com/
|
|
| Sep 21, 2016 at 13:18 | comment | added | Christoph Hanck | See the next edit! | |
| Sep 21, 2016 at 13:18 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 323 characters in body
|
| Sep 21, 2016 at 13:12 | comment | added | Michal | @ Christoph Hanck you are right, I must have made something incorrect with the code, I have rerun from scratch and it returns the correct values. How about the 3. case (with constan and a trend) how the equation would have to be adjusted for it? | |
| Sep 21, 2016 at 12:53 | comment | added | Christoph Hanck |
You are right about the right critical values, but for me it does produce these values: > (CriticalValues <- sort(DFstats)[c(0.01,0.05,0.1)*reps]) yields [1] -3.412332 -2.868623 -2.569988
|
|
| Sep 21, 2016 at 12:45 | comment | added | Michal | small correction for T=500 reps=50000 it should be -3.43 -2.86 -2.57 | |
| Sep 21, 2016 at 12:29 | comment | added | Michal | @ Christoph Hanck using the code in the edit section for the 2. case with constant and without trend doesn't give the correct results, for example for T=500 reps=50000 it should be something like -3.45 -2.87 -2.58 for 1% 5% 10% but it returns much lower values (for 50000 simulations -4.35 -3.897 -3.68) | |
| Sep 21, 2016 at 11:11 | comment | added | Christoph Hanck | I am sorry, the edit was of course for the case with constant and no trend, not with trend. I corrected that. I also tried to further explain the code. | |
| Sep 21, 2016 at 11:10 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 243 characters in body
|
| Sep 21, 2016 at 10:46 | comment | added | Michal | @ Christoph Hanck, my understanding is that the code in the edition is for the case with trend and without constant, how about the case 2. without trend and with a constant and 3. with both constant and trend. I can't get the logic of those adjustments, could you please clarify the two last lines of the loop? | |
| Sep 21, 2016 at 6:38 | comment | added | Christoph Hanck | @Michal, OK, see the edit. | |
| Sep 21, 2016 at 6:38 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 535 characters in body
|
| Sep 20, 2016 at 19:41 | comment | added | Michal | @ Christoph Hanck the above is the 1. type with no constant and without trend. what adjustment would have to be made to do the two other types of ADF test (2.with a constant and without trend 3. with a constant and a trend) | |
| Jun 25, 2016 at 12:30 | vote | accept | Michal | ||
| May 20, 2016 at 10:51 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 46 characters in body
|
| May 20, 2016 at 9:30 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 1 character in body
|
| May 20, 2016 at 9:21 | history | edited | Christoph Hanck | CC BY-SA 3.0 |
added 1309 characters in body
|
| May 20, 2016 at 8:38 | history | answered | Christoph Hanck | CC BY-SA 3.0 |