rtadf
Moderators: EViews Gareth, EViews Moderator, EViews Esther
Re: rtadf
Hello.
Unfortunately I am having exactly the same problem with version 1.4. After running the GSADF test is says 'BETA_PC' not defined and it doesn't produce the gsadf_cv series. Unfortunately I am quickly running out of time to submit my dissertation so I would appreciate any help!
Many thanks!
Unfortunately I am having exactly the same problem with version 1.4. After running the GSADF test is says 'BETA_PC' not defined and it doesn't produce the gsadf_cv series. Unfortunately I am quickly running out of time to submit my dissertation so I would appreciate any help!
Many thanks!
Re: rtadf
A new version of the Rtadf addin has been upladed to the EViews Addins repository (http://www.eviews.com/Addins/addins.shtml#addins).
The main improvement added to the current version is the option to simulate critical values for the SADF and GSADF tests in parallel mode via the use (in the background) of MATLAB and the Parallel Computation Toolbox. This new feature significantly decreases running time, especially for the GSADF test.
A few notes:
1. The addin includes an updated user guide.
2. This version is still in beta stages, so it might include some unresolved bugs. If you do get the chance to experiment with it a little, I'll be more than happy if you'll send me some feedback (That annoying "BETA_PC bug" is hopfully fixed).
3. In order to use parallel simulation option, you must have MATLAB + Parallel Computation Toolbox installed on your machine.
Best,
Itamar
The main improvement added to the current version is the option to simulate critical values for the SADF and GSADF tests in parallel mode via the use (in the background) of MATLAB and the Parallel Computation Toolbox. This new feature significantly decreases running time, especially for the GSADF test.
A few notes:
1. The addin includes an updated user guide.
2. This version is still in beta stages, so it might include some unresolved bugs. If you do get the chance to experiment with it a little, I'll be more than happy if you'll send me some feedback (That annoying "BETA_PC bug" is hopfully fixed).
3. In order to use parallel simulation option, you must have MATLAB + Parallel Computation Toolbox installed on your machine.
Best,
Itamar
Re: rtadf
Dear bm26,
Thanks! Unfortunately, what you ask for is not possible in the current version. However, If I understand you correctly, I think you can get a backward GSADF test if you somehow "rotate" you series such that the last observation becomes the first and then run the test, no?
Thanks! Unfortunately, what you ask for is not possible in the current version. However, If I understand you correctly, I think you can get a backward GSADF test if you somehow "rotate" you series such that the last observation becomes the first and then run the test, no?
Re: rtadf
Dear Itamar,
if I understand correctly, the "windows" of your test are moving from the left to the right of the considered time series. Phillips,Shi and Yu suggest in a later version of their paper that a rolling window from the right to the left might be better. I think that a "turnaround" of my time series is technically a possibe way, but I fear that I cannot present the output, because the time series is then of course incorrect.
if I understand correctly, the "windows" of your test are moving from the left to the right of the considered time series. Phillips,Shi and Yu suggest in a later version of their paper that a rolling window from the right to the left might be better. I think that a "turnaround" of my time series is technically a possibe way, but I fear that I cannot present the output, because the time series is then of course incorrect.

 Posts: 1
 Joined: Thu Jul 14, 2016 4:12 pm
Re: rtadf
Hey guys i just ran my model today using the addin.. and i was wondering how exactly can we do see the datestamping, i am a bit new to this.
Thank you in advance.
Vasileios
Thank you in advance.
Vasileios
Re: rtadf
Hi,
In short, a bubble period is a period when the ADF sequence is larger than the ADF critical value sequence. For example, if you run the sadf test, your workfile will include two new series  "sadf" and sadf_cv". You can generate a bubble period dummy indicator by running something like
Note that there is an issue of a minimal threshold for a bubble period. See Phillips et al. (2016).
In short, a bubble period is a period when the ADF sequence is larger than the ADF critical value sequence. For example, if you run the sadf test, your workfile will include two new series  "sadf" and sadf_cv". You can generate a bubble period dummy indicator by running something like
Code: Select all
bubdum=@recode(sadf>sadf_cv,1,0)
Note that there is an issue of a minimal threshold for a bubble period. See Phillips et al. (2016).
Re: rtadf
Dear Itamar,
Thanks for a great addin! Without your work I wouldn't be able to write my MSc thesis so smoothly. Being a basic Eviews user I rested solely on the visual interface you created (i.e. I wasn't using command line at all) and enjoyed it much. However, I have found some miscrepancies you may like to correct in the next updates together with a new functionality I would love to see. Please treat these comments as a feedback from a grateful user who wants to help future generations.
1. [fix] When choosing lag lenght selection criteria the interface allows to choose one of a few of criteria but the output table always shows a wrong criterion  the addin prints "minus 1" (in terms of the occurence order) criterion. I have found a selfmade tweak for this proble  in the addin code: if !criterion = 2 should be changed to 3, if !criterion = 3 should be changed to 4 etc. However, there might be a better fix I'm not aware of.
2. [fix] There is a number of typos in the addin. Correcting them will surely enhance the user experience and make the addin more straightforward in usage. To name a few:
'and requiers the Parallel Computing Toolbox' > 'and requires the Parallel Computing Toolbox';
Foward ADF sequence (left axis) > Forward ADF sequence (left axis);
Bacwards SADF sequence (left axis) > Backwards SADF sequence (left axis).
3. [new function] In my thesis I have used many same lenght series (with very similar characteristics) to test the bubble existence hypothesis with SADF & GSADF tests. As a basic Eviews user I needed to run the test together with the simulation over and over again even though I knew that the results of the simulation are the same. Running the SADF test simulation a number of times was managable timewise, but in the case of GSADF test the simulation was so sluggish I had to come up with a slight modification/tweak of the addin code just to print the graphs I needed and omit the time consuming simulation. If it's not complicated to implement, I would greatly appreciate a small checkbox in your addin which will allow to print the results of the test using critical values obtained from the previous simulation (i.e. "obtain critical values from the previous simulation").
Hope this will help develop the addin!
Thanks for a great addin! Without your work I wouldn't be able to write my MSc thesis so smoothly. Being a basic Eviews user I rested solely on the visual interface you created (i.e. I wasn't using command line at all) and enjoyed it much. However, I have found some miscrepancies you may like to correct in the next updates together with a new functionality I would love to see. Please treat these comments as a feedback from a grateful user who wants to help future generations.
1. [fix] When choosing lag lenght selection criteria the interface allows to choose one of a few of criteria but the output table always shows a wrong criterion  the addin prints "minus 1" (in terms of the occurence order) criterion. I have found a selfmade tweak for this proble  in the addin code: if !criterion = 2 should be changed to 3, if !criterion = 3 should be changed to 4 etc. However, there might be a better fix I'm not aware of.
2. [fix] There is a number of typos in the addin. Correcting them will surely enhance the user experience and make the addin more straightforward in usage. To name a few:
'and requiers the Parallel Computing Toolbox' > 'and requires the Parallel Computing Toolbox';
Foward ADF sequence (left axis) > Forward ADF sequence (left axis);
Bacwards SADF sequence (left axis) > Backwards SADF sequence (left axis).
3. [new function] In my thesis I have used many same lenght series (with very similar characteristics) to test the bubble existence hypothesis with SADF & GSADF tests. As a basic Eviews user I needed to run the test together with the simulation over and over again even though I knew that the results of the simulation are the same. Running the SADF test simulation a number of times was managable timewise, but in the case of GSADF test the simulation was so sluggish I had to come up with a slight modification/tweak of the addin code just to print the graphs I needed and omit the time consuming simulation. If it's not complicated to implement, I would greatly appreciate a small checkbox in your addin which will allow to print the results of the test using critical values obtained from the previous simulation (i.e. "obtain critical values from the previous simulation").
Hope this will help develop the addin!
Re: rtadf
Hi,
FYI, a new version of the Rtadf addin has been uploaded to the EViews addin repository. The main improvement being added is the option to simulate critical values for the SADF and GSADF tests based on wild and sieve bootstrap (as in as in Harvey et al. (2016) and Gutierrez (2011)).
A few notes:
1. All the suggestions made by aglonger (in the above post) are implemented, including the option to conduct statistical inference based on a previously run simulation. (BTW, I would like to thank aglonger for his wonderful suggestions. Your post was the reason I finally sat down and worked on this update!)
2. The addin includes a draft of an updated and detailed user guide.
3. This version is still in beta stages, so it might include some unresolved bugs.
4. The bootstrap simulation is currently only available in EViews. Soon (hopefully), I will add the option to use MATLAB + Parallel Computation Toolbox (as in the Monte Carlo based simulation).
I'll be more than happy to receive any feedback, comments or suggestions you may have.
Best,
Itamar
References
Gutierrez, L. (2011). Bootstrapping asset price bubbles. Economic Modelling, 28(6), 2488–2493.
Harvey, D. I., Leybourne, S. J., Sollis, R., & Taylor, A. M. R. (2016). Tests for explosive financial bubbles in the presence of nonstationary volatility. Journal of Empirical Finance, 38, 548–574.
FYI, a new version of the Rtadf addin has been uploaded to the EViews addin repository. The main improvement being added is the option to simulate critical values for the SADF and GSADF tests based on wild and sieve bootstrap (as in as in Harvey et al. (2016) and Gutierrez (2011)).
A few notes:
1. All the suggestions made by aglonger (in the above post) are implemented, including the option to conduct statistical inference based on a previously run simulation. (BTW, I would like to thank aglonger for his wonderful suggestions. Your post was the reason I finally sat down and worked on this update!)
2. The addin includes a draft of an updated and detailed user guide.
3. This version is still in beta stages, so it might include some unresolved bugs.
4. The bootstrap simulation is currently only available in EViews. Soon (hopefully), I will add the option to use MATLAB + Parallel Computation Toolbox (as in the Monte Carlo based simulation).
I'll be more than happy to receive any feedback, comments or suggestions you may have.
Best,
Itamar
References
Gutierrez, L. (2011). Bootstrapping asset price bubbles. Economic Modelling, 28(6), 2488–2493.
Harvey, D. I., Leybourne, S. J., Sollis, R., & Taylor, A. M. R. (2016). Tests for explosive financial bubbles in the presence of nonstationary volatility. Journal of Empirical Finance, 38, 548–574.

 Posts: 1
 Joined: Fri Dec 02, 2016 5:19 pm
Re: rtadf
Hi Itamar,
First of all, thank you for the addin! It saved my thesis on housing bubbles.
I have some questions according to PSY GSADF Method in the addin and the critical value.
1, I am confused about the test 4 gsadf. Are you saying that the gsadf here is the sup bsadf test PSY(2012) used in the paper?
2, I set aic as lag length, 0 lag, monte carlo eviews, 1000 replication, beta 0.95, d=1, eta=1. After I run the test, he graph shows me 0 for the 95%CV for every test I run. However, in the table the tstatistic does for 95% CV is not 0. Is my set up wrong? Or do I need to draw the CV line by myself?
Thank you!
First of all, thank you for the addin! It saved my thesis on housing bubbles.
I have some questions according to PSY GSADF Method in the addin and the critical value.
1, I am confused about the test 4 gsadf. Are you saying that the gsadf here is the sup bsadf test PSY(2012) used in the paper?
2, I set aic as lag length, 0 lag, monte carlo eviews, 1000 replication, beta 0.95, d=1, eta=1. After I run the test, he graph shows me 0 for the 95%CV for every test I run. However, in the table the tstatistic does for 95% CV is not 0. Is my set up wrong? Or do I need to draw the CV line by myself?
Thank you!
Re: rtadf
Hi francescao,
Thanks!
As for your questions:
1. The GSADF is the sup of the BSADF used in Phillips, Shi and Yu (2015, IER)
2. The CV in the table is for the sup test, whereas the red line is the CV for the datestamping procedure. You can checkout PSY (2015) or the addins' user guide for elaborated explanations.
Good luck
Thanks!
As for your questions:
1. The GSADF is the sup of the BSADF used in Phillips, Shi and Yu (2015, IER)
2. The CV in the table is for the sup test, whereas the red line is the CV for the datestamping procedure. You can checkout PSY (2015) or the addins' user guide for elaborated explanations.
Good luck
Re: rtadf
Hi Itamar!
I've recently been trying to use the rtadf Addin, but i couldn't succesfuly finish the estimation of the GSADF (PSY 2015) test.
Everything goes well until the critical values are being estimated (Monte Carlo Matlab option selected). This is the error message i get every time: "TSTAT_MAX_MAX is not defined"
Could you give me further guidance on how to solve it?
Thank you,
Julian
I've recently been trying to use the rtadf Addin, but i couldn't succesfuly finish the estimation of the GSADF (PSY 2015) test.
Everything goes well until the critical values are being estimated (Monte Carlo Matlab option selected). This is the error message i get every time: "TSTAT_MAX_MAX is not defined"
Could you give me further guidance on how to solve it?
Thank you,
Julian
 Attachments

 5.png (106.05 KiB) Viewed 2387 times
Who is online
Users browsing this forum: No registered users and 3 guests