Roll (Rolling Regression)
Moderators: EViews Gareth, EViews Moderator, EViews Esther

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
They're stored in your workfile under the name "rollcoefs_"
Follow us on Twitter @IHSEViews

 Posts: 1
 Joined: Mon Aug 22, 2011 1:23 pm
Re: Roll (Rolling Regression)
Hi all.
I´m trying to perform a rolling regression with a panel (42 periods x 4 crosssections) but I get this message: "Error in Sample: Range Error".
I´ve tried modifying the range and the sample but I can´t make it work.
I´m a little rusty with my EV (I started with the 3.1 version back at school...). Any ideas?
Thanks in advance!
I´m trying to perform a rolling regression with a panel (42 periods x 4 crosssections) but I get this message: "Error in Sample: Range Error".
I´ve tried modifying the range and the sample but I can´t make it work.
I´m a little rusty with my EV (I started with the 3.1 version back at school...). Any ideas?
Thanks in advance!

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
The addin doesn't support Panel data (sorry!).
Follow us on Twitter @IHSEViews
Re: Roll (Rolling Regression)
Hello,
i think there's a bug in the addin:
if i do a rolling regression which includes more than 1 ar() / ma() parameter, i get this error: "Invalid (or out of range) coefficient or matrix index 2."
the thing is that, one i use only 1 ar() or 1 ma() parameter the addin works but the coefficients are stored in a series called rollcoefs_ instead of being called, for example, rollcoefs_ar1
I'm currently looking at the source code in order to fix this behaviour but if someone already knows how to fix it i would defintely appreciate it!
i think there's a bug in the addin:
if i do a rolling regression which includes more than 1 ar() / ma() parameter, i get this error: "Invalid (or out of range) coefficient or matrix index 2."
the thing is that, one i use only 1 ar() or 1 ma() parameter the addin works but the coefficients are stored in a series called rollcoefs_ instead of being called, for example, rollcoefs_ar1
I'm currently looking at the source code in order to fix this behaviour but if someone already knows how to fix it i would defintely appreciate it!

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
A real quick fix is to change this bit of code (in the roll.prg):
to be just (i.e. get rid of the if condition completely):
Code: Select all
if _this.@bylist=1 then
%coefnames = _this.@varlist
%coefnames = @wmid(%coefnames,2) 'drop dependent var.
else
%coefnames = _this.@coeflist
endif
to be just (i.e. get rid of the if condition completely):
Code: Select all
%coefnames = _this.@coeflist
Follow us on Twitter @IHSEViews
Re: Roll (Rolling Regression)
hi,
Im new in this forum.. Im running an AR(1) process, I estimate a rolling sample, window fo 20,
How can i get the standard deviation or regression residuals, and a Estimated Aerage of de dependent variable ??
im using the roll add in.
thanks,
Im new in this forum.. Im running an AR(1) process, I estimate a rolling sample, window fo 20,
How can i get the standard deviation or regression residuals, and a Estimated Aerage of de dependent variable ??
im using the roll add in.
thanks,
Re: Roll (Rolling Regression)
Hi,
I am using ADVROLL to run a rolling regression.
My window size is set to 90 and my step is set to 1.
After I run the rolling regression, my first coefficients are at observation 90.
I have NA for the first 89 observation points which makes sense because of the window size I have chosen.
I also get NA for the last observation point in my series.
All of my coefficients for the final date in the series are NA.
Why is this the case?
Thanks,
Jen
I am using ADVROLL to run a rolling regression.
My window size is set to 90 and my step is set to 1.
After I run the rolling regression, my first coefficients are at observation 90.
I have NA for the first 89 observation points which makes sense because of the window size I have chosen.
I also get NA for the last observation point in my series.
All of my coefficients for the final date in the series are NA.
Why is this the case?
Thanks,
Jen

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
I have attached the workfile, thanks for taking a look.
 Attachments

 elr_returns.wf1
 (213.94 KiB) Downloaded 244 times

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
Works fine for me. Try redownloading the addin.
Follow us on Twitter @IHSEViews
Re: Roll (Rolling Regression)
Hello
Can someone guide me on how to store the S.E. of regressions while using the rolling regression addin? Thanks in advance.
Can someone guide me on how to store the S.E. of regressions while using the rolling regression addin? Thanks in advance.

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
You'll have to add it to the program yourself, I think. Find the code that stores the Rsquareds, and either change it to store @se rather than @r2, or add new code to do the SE.
Follow us on Twitter @IHSEViews
Re: Roll (Rolling Regression)
Hi,
I'm getting such an error when I try to install the addin.
"ADDIN is not defined or is an illegal command in...". Is it because I'm using an inferior version of eviews or should I download something else in advance.
Thanks,
ps. my version of eviews is eviews7.
I'm getting such an error when I try to install the addin.
"ADDIN is not defined or is an illegal command in...". Is it because I'm using an inferior version of eviews or should I download something else in advance.
Thanks,
ps. my version of eviews is eviews7.

 Fe ddaethom, fe welon, fe amcangyfrifon
 Posts: 11983
 Joined: Tue Sep 16, 2008 5:38 pm
Re: Roll (Rolling Regression)
I have installed eviews 7.2 version. I have been running a rolling regression on the previous version, a 30week 1 step roll regression. In the new version, when I run the regression the same way my latest week data gives me NA as coefficients for all the numbers. Do you know what this is about? Also, can I force the intercept as zero?
Who is online
Users browsing this forum: No registered users and 5 guests