Page 2 of 2
Re: time varying autoregressive process with stochastic vola
Posted: Tue May 08, 2018 8:44 am
by TomDoan
That sounds like it's the same idea as Primiceri's but instead of a long IRF at several locations, it's a short IRF at a lot of locations. Does that sound correct?
Re: time varying autoregressive process with stochastic vola
Posted: Thu May 10, 2018 9:47 am
by BinhPham
Hi Tom,
Fig 4 and 5 of Primiceri (2005) are computed by computing irf, say 60 periods horizon, for each point of time. Then plot irf(0), irf(10), irf(20) and irf(60) over time where irf(.) denotes response value at h=0, 10, 20 and 60, aren't they?. So, I would loop from t=1 to t=T, calculate irf at each t; collecting them in four series! Is this correct?
Another question, in the paper Primiceri calculated 1% permanent impulse-response of inflation and unrate. What does exactly it mean? I am confusing with "permanent increase" since we have structural shock e_inf = 1 at time t=0 and zeros otherwise. Could you please point me out here?
Many thanks.
insufficient memory
Posted: Sat Mar 16, 2019 4:25 am
by pascal
Dear all,
I am trying to use time varying VAR for estimation with daily data. There are 4 variables in the system. However, It shows the following message:
## M4. A memory request for an additional 199148820 bytes cannot be satisfied
The Error Occurred At Location 6257, Line 396 of VARTVPKSC
C:\Users\User\Desktop\Time Varying Structural Vector Autoregressions\VARTVPKSC.src Line 477
May I know how can I fix this problem?
Many thanks
Pascal
Re: insufficient memory
Posted: Tue Mar 19, 2019 10:35 am
by TomDoan
pascal wrote:Dear all,
I am trying to use time varying VAR for estimation with daily data. There are 4 variables in the system. However, It shows the following message:
## M4. A memory request for an additional 199148820 bytes cannot be satisfied
The Error Occurred At Location 6257, Line 396 of VARTVPKSC
C:\Users\User\Desktop\Time Varying Structural Vector Autoregressions\VARTVPKSC.src Line 477
May I know how can I fix this problem?
Many thanks
Pascal
That sounds like it's a rather large problem for this type of analysis (which is generally done with macro data, not financial). The calculations require quite a bit of storage space as they need to keep responses for each draw and for each time period in the estimation range (since the coefficients are time-varying). If you cut down on the number of draws, it will reduce the storage requirements proportionally to the number of draws that you take. WinRATS Pro also has effectively unlimited storage on the 64-bit application, while an application in WinRATS Standard is limited to 2Gb.
Re: time varying autoregressive process with stochastic vola
Posted: Mon Mar 25, 2019 11:17 am
by pascal
Hi Tom,
The problem "A memory request for an additional 199148820 bytes cannot be satisfied" has been solved. Thank you for your help.
May I know how should I do if am getting the following error:
## MAT14. Non-invertible Matrix. Using Generalized Inverse for SYMMETRIC.
The Error Occurred At Location 311, Line 15 of loop/block
I am a beginner in this software. Sorry for asking you many questions.
Thank you
Pascal
Re: time varying autoregressive process with stochastic vola
Posted: Mon Mar 25, 2019 6:50 pm
by TomDoan
That seems to be a known problem. See another thread starting at
https://estima.com/forum/viewtopic.php?p=13771#p13771
Re: time varying autoregressive process with stochastic vola
Posted: Tue Mar 26, 2019 1:11 am
by pascal
Dear Tom,
Thank you for your kindly help.
Many thanks.
Pascal