2:58 p.m.
I wanted to estimate a model with a time varying parameter (on a simple linear trend) and
would appreciate some guidance as I am not a programmer in Gretl's language.
1) Where in the GUI Model set of commands do you access the State Space modelling - is by
writing the spec in the MLE option?
2) I loaded the example file AWM.gdt
3)Entered the commands below exactly as specified in the manual viz. Listing 34.4:
Phillips curve on Euro data with time-varying slope
4) It died at the first line.
function void at_each_step(bundle *b)
b.obsymat = transp(b.mX[b.t,])
end function
open AWM.gdt --quiet
smpl 1974:1 1994:1
/* parameter initialization */
scalar b0 = mean(INFQ)
scalar s_obs = 0.1
scalar s_state = 0.1
/* bundle setup */
bundle B = ksetup(INFQ, 1, 1, 1)
matrix B.mX = {URX}
matrix B.depvar = {INFQ}
B.timevar_call = "at_each_step"
B.diffuse = 1
/* ML estimation of intercept and the two variances */
mle LL = err ? NA : B.llt
B.obsy = B.depvar - b0
B.obsvar = s_obs^2
B.statevar = s_state^2
err = kfilter(&B)
params b0 s_obs s_state
end mle
/* display the smoothed time-varying slope */
ksmooth(&B)
series tvar_b1hat = B.state[,1]
series tvar_b1se = sqrt(B.stvar[,1])
gnuplot tvar_b1hat --time-series --with-lines --output=display \
--band=tvar_b1hat,tvar_b1se,1.96 --band-style=fill
3:52 p.m.
Am 01.12.2020 um 15:58 schrieb bjr.newmail(a)gmail.com:
I wanted to estimate a model with a time varying parameter (on a
simple linear trend) and would appreciate some guidance as I am not a programmer in
Gretl's language.
1) Where in the GUI Model set of commands do you access the State Space modelling - is by
writing the spec in the MLE option?
2) I loaded the example file AWM.gdt
3)Entered the commands below exactly as specified in the manual viz. Listing 34.4:
Phillips curve on Euro data with time-varying slope
4) It died at the first line.
Hi, I copied and pasted the code from your email and it runs fine here.
Please provide more details about the error message you're getting and
your gretl version etc.
As for your general question: AFAIK there is currently no GUI access to
the general state-space framework yet, given the complexity involved.
For certain special cases the contributed package StrucTiSM offers a GUI
access, among which the local level, the local linear trend, and the
so-called basic structural models. Don't know if it really covers what
you're looking for, though.
cheers
sven
4:20 p.m.
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In the
Model MLE option from the GUI -or under File -Script File-menu option? As
I said I have used Gretl so far only via the GUI. -so where do I go to
paste it in and run the problem from my data file. I think I can alter the
example variables and parameters. It is just to know where to do this. The
manual is not helpful in this respect. I suspect the error message will
not be of much help if I was inserting the Script for the SSmodel in the
wrong place.
cheers
Brian
On Tue, 1 Dec 2020 at 15:52, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 15:58 schrieb bjr.newmail(a)gmail.com:
> I wanted to estimate a model with a time varying parameter (on a simple
linear trend) and would appreciate some guidance as I am not a programmer
in Gretl's language.
> 1) Where in the GUI Model set of commands do you access the State Space
modelling - is by writing the spec in the MLE option?
> 2) I loaded the example file AWM.gdt
> 3)Entered the commands below exactly as specified in the manual viz.
Listing 34.4: Phillips curve on Euro data with time-varying slope
> 4) It died at the first line.
>
Hi, I copied and pasted the code from your email and it runs fine here.
Please provide more details about the error message you're getting and
your gretl version etc.
As for your general question: AFAIK there is currently no GUI access to
the general state-space framework yet, given the complexity involved.
For certain special cases the contributed package StrucTiSM offers a GUI
access, among which the local level, the local linear trend, and the
so-called basic structural models. Don't know if it really covers what
you're looking for, though.
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
5:10 p.m.
Am 01.12.2020 um 17:20 schrieb Brian Revell:
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In
the Model MLE option from the GUI -or under File -Script File-menu
option?
The latter, yes, so basically in the built-in script editor. The MLE
area is meant for even lower-level access, without the predefined Kalman
filter objects.
know where to do this. The manual is not helpful in this respect.
In later parts of the guide it is pretty much assumed that people know
the scripting interface of gretl. However, at the beginning this is
explained, please see section 3.1.
(BTW, in my version of the guide the script listing is number 36.4 in
ch. 36, I guess you have a slightly older version where some newer
chapters do not exist yet. But this shouldn't make a difference in this
context.)
cheers
sven
5:14 p.m.
Thanks again. Hopefully it will ho smoothly
Cheers
Brian
On Tue, 1 Dec 2020, 17:10 Sven Schreiber, <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 17:20 schrieb Brian Revell:
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In the
Model MLE option from the GUI -or under File -Script File-menu option?
The latter, yes, so basically in the built-in script editor. The MLE area
is meant for even lower-level access, without the predefined Kalman filter
objects.
know where to do this. The manual is not helpful in this respect.
In later parts of the guide it is pretty much assumed that people know the
scripting interface of gretl. However, at the beginning this is explained,
please see section 3.1.
(BTW, in my version of the guide the script listing is number 36.4 in ch.
36, I guess you have a slightly older version where some newer chapters do
not exist yet. But this shouldn't make a difference in this context.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
3:29 p.m.
Hi again Sven
perhaps I am being dim and most probably a pain! . Basically I am
interested determining the time varying trend parameter in a simple trend
relationship as in 36.4
[image: image.png]
where in the Script below y =Severnadj and x = Year. I created and
entered a new variable Year as the time trend time variable generated by
Gretl failed to work in an earlier run.
However MLE fails to recognize Year in the script entered. The datafile
loads correctly, mean of Severadj is correct.
Where am I going wrong.?
Script entered.
__________________________________________
function void at_each_step(bundle *b)
b.obsymat = transp(b.mX[b.t,])
end function
open WELSHEGGSCL.gdt --quiet
smpl 2002 2019
/* parameter initialization */
scalar b0 = mean(Severnadj)
scalar s_obs = 0.1
scalar s_state = 0.1
/* bundle setup */
bundle B = ksetup(Severnadj, 1, 1, 1)
matrix B.mX = {Year}
matrix B.depvar = {Severnadj}
B.timevar_call = "at_each_step"
B.diffuse = 1
/* ML estimation of intercept and the two variances */
mle LL = err ? NA : B.llt
B.obsy = B.depvar - b0
B.obsvar = s_obs^2
B.statevar = s_state^2
err = kfilter(&B)
params b0 s_obs s_state
end mle
/* display the smoothed time-varying slope */
ksmooth(&B)
series tvar_b1hat = B.state[,1]
series tvar_b1se = sqrt(B.stvar[,1])
gnuplot tvar_b1hat --time-series --with-lines --output=display \
--band=tvar_b1hat,tvar_b1se,1.96 --band-style=fill
___________________________________________________________
Final error messages from log :-
/* parameter initialization */
? scalar b0 = mean(Severnadj)
Generated scalar b0 = 9.84213
? scalar s_obs = 0.1
Generated scalar s_obs = 0.1
? scalar s_state = 0.1
Generated scalar s_state = 0.1
/* bundle setup */
? bundle B = ksetup(Severnadj, 1, 1, 1)
? matrix B.mX = {Year}
B.mX = {Year}
The symbol 'Year' is undefined
Error executing script: halting
matrix B.mX = {Year}
cheers
Brian
On Tue, 1 Dec 2020 at 17:10, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 17:20 schrieb Brian Revell:
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In the
Model MLE option from the GUI -or under File -Script File-menu option?
The latter, yes, so basically in the built-in script editor. The MLE area
is meant for even lower-level access, without the predefined Kalman filter
objects.
know where to do this. The manual is not helpful in this respect.
In later parts of the guide it is pretty much assumed that people know the
scripting interface of gretl. However, at the beginning this is explained,
please see section 3.1.
(BTW, in my version of the guide the script listing is number 36.4 in ch.
36, I guess you have a slightly older version where some newer chapters do
not exist yet. But this shouldn't make a difference in this context.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
3:37 p.m.
PS -I should say that the demo script in the User Guide ran perfectly, so
the problem seems to be in what I have changed -viz filename, dep var
Severnadj and Indep var Year.
Brian
On Tue, 1 Dec 2020 at 17:10, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 17:20 schrieb Brian Revell:
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In the
Model MLE option from the GUI -or under File -Script File-menu option?
The latter, yes, so basically in the built-in script editor. The MLE area
is meant for even lower-level access, without the predefined Kalman filter
objects.
know where to do this. The manual is not helpful in this respect.
In later parts of the guide it is pretty much assumed that people know the
scripting interface of gretl. However, at the beginning this is explained,
please see section 3.1.
(BTW, in my version of the guide the script listing is number 36.4 in ch.
36, I guess you have a slightly older version where some newer chapters do
not exist yet. But this shouldn't make a difference in this context.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
4:04 p.m.
Hi Sven
progress - I have managed to get the Script to run with *time *as indep
var -but MLE crashes on reaching max 500 interations . It didn't like the
year variable I defined in an earlier run eg 2002, 2003 I think because
it was a large magnitude different eg 10^3 to the dependent variable.
So is there a way to increase the max no. of iterations in the MLE
routine? see below for output..
? function void at_each_step(bundle *b)
b.obsymat = transp(b.mX[b.t,])
end function
? open WELSHEGGSCL.gdt --quiet
Read datafile C:\Users\Brian\Documents\gretl\WELSHEGGSCL.gdt
? smpl 2002 2019
Full data range: 2002 - 2024 (n = 23)
Current sample: 2002 - 2019 (n = 18)
/* parameter initialization */
? scalar b0 = mean(Severnadj)
Generated scalar b0 = 9.84213
? scalar s_obs = 0.1
Generated scalar s_obs = 0.1
? scalar s_state = 0.1
Generated scalar s_state = 0.1
/* bundle setup */
? bundle B = ksetup(Severnadj, 1, 1, 1)
? matrix B.mX = {time}
? matrix B.depvar = {Severnadj}
? B.timevar_call = "at_each_step"
? B.diffuse = 1
/* ML estimation of intercept and the two variances */
? mle LL = err ? NA : B.llt
? B.obsy = B.depvar - b0
? B.obsvar = s_obs^2
? B.statevar = s_state^2
? err = kfilter(&B)
? params b0 s_obs s_state
? end mle
Using numerical derivatives
Tolerance = 1.81899e-012
*Reached the maximum iterations (500)*
Error executing script: halting
end mle
On Tue, 1 Dec 2020 at 17:10, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 17:20 schrieb Brian Revell:
Hi Sven
many thanks for your prompt feedback. Where did you paste it in? In the
Model MLE option from the GUI -or under File -Script File-menu option?
The latter, yes, so basically in the built-in script editor. The MLE area
is meant for even lower-level access, without the predefined Kalman filter
objects.
know where to do this. The manual is not helpful in this respect.
In later parts of the guide it is pretty much assumed that people know the
scripting interface of gretl. However, at the beginning this is explained,
please see section 3.1.
(BTW, in my version of the guide the script listing is number 36.4 in ch.
36, I guess you have a slightly older version where some newer chapters do
not exist yet. But this shouldn't make a difference in this context.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
5:43 p.m.
Am 02.12.2020 um 17:04 schrieb Brian Revell:
Hi Sven
progress - I have managed to get the Script to run with *time *as
indep var
Yes, previously you said you had created the variable "Year",
but it was
unknown to gretl so I'm not sure what you did. In a time-series dataset
you can use the generic predefined special variable "time" as you did,
or if you want to really aim for the year for example also in a
quarterly or monthly dataset you could grab the series dollar-style
accessor "$obsmajor" (see the function reference for the dollar accessors).
So is there a way to increase the max no. of iterations in the MLE
routine?
In principle you can manipulate that via a line such as "set
bfgs_maxiter 1000". See the doc for the "set" command. The background is
that MLE in gretl uses the BFGS optimizer (don't know right now if this
is mentioned in the guide, but it is in the "set" help).
However...
Full data range: 2002 - 2024 (n = 23)
Current sample: 2002 - 2019 (n = 18)
you have a pretty small sample. This is probably not large enough to
work with time-varying stuff and expect reasonably identified results.
This could also explain the convergence problems (among other reasons).
So I wouldn't be surprised if increasing the iterations doesn't help.
But you might as well try.
cheers
sven
7:26 p.m.
Hi Sven
with the Script below which is for a time varying parameter. there are 2
things I wish to get from the output of this routine.
First in the final gnuplot command I have added, I want to include in the
graph of the dependent variable the estimated/smoothed level of the
system with its 95% CI
Second to get a printout of the values of tvar_ bhat and its 95% CI and
of the level of the system
Also with regard to StrucTiSM, it would be helpful have a printout of the
values of estimated level of the system (and its 95%CI)
Is there some simple script I can add to do these things.? The forecast
script you supplied works well thank you in StrucTism post estimation.
cheers
Brian
_________________________________________________________-
function void at_each_step(bundle *b)
b.obsymat = transp(b.mX[b.t,])
end function
open SevernDeeEggs.gdt --quiet
smpl 1998 2019
/* parameter initialization */
scalar b0 = mean(SEVERN_EGGS_UNADJ)
scalar s_obs = 0.1
scalar s_state = 0.1
/* bundle setup */
bundle B = ksetup(SEVERN_EGGS_UNADJ, 1, 1, 1)
matrix B.mX = {time}
matrix B.depvar = {SEVERN_EGGS_UNADJ}
B.timevar_call = "at_each_step"
B.diffuse = 1
set bfgs_maxiter 5000
/* ML estimation of intercept and the two variances */
mle LL = err ? NA : B.llt
B.obsy = B.depvar - b0
B.obsvar = s_obs^2
B.statevar = s_state^2
err = kfilter(&B)
params b0 s_obs s_state
end mle
/* display the smoothed time-varying slope */
ksmooth(&B)
series tvar_b1hat = B.state[,1]
series tvar_b1se = sqrt(B.stvar[,1])
gnuplot tvar_b1hat --time-series --with-lines --output=display \
--band=tvar_b1hat,tvar_b1se,1.96 --band-style=fill
*gnuplot SEVERN_EGGS_UNADJ --time-series --with-lines --output=display*
____________________________________________________________________
On Wed, 2 Dec 2020 at 16:04, Brian Revell <bjr.newmail(a)gmail.com> wrote:
Hi Sven
progress - I have managed to get the Script to run with *time *as indep
var -but MLE crashes on reaching max 500 interations . It didn't like the
year variable I defined in an earlier run eg 2002, 2003 I think because
it was a large magnitude different eg 10^3 to the dependent variable.
So is there a way to increase the max no. of iterations in the MLE
routine? see below for output..
? function void at_each_step(bundle *b)
> b.obsymat = transp(b.mX[b.t,])
> end function
? open WELSHEGGSCL.gdt --quiet
Read datafile C:\Users\Brian\Documents\gretl\WELSHEGGSCL.gdt
? smpl 2002 2019
Full data range: 2002 - 2024 (n = 23)
Current sample: 2002 - 2019 (n = 18)
/* parameter initialization */
? scalar b0 = mean(Severnadj)
Generated scalar b0 = 9.84213
? scalar s_obs = 0.1
Generated scalar s_obs = 0.1
? scalar s_state = 0.1
Generated scalar s_state = 0.1
/* bundle setup */
? bundle B = ksetup(Severnadj, 1, 1, 1)
? matrix B.mX = {time}
? matrix B.depvar = {Severnadj}
? B.timevar_call = "at_each_step"
? B.diffuse = 1
/* ML estimation of intercept and the two variances */
? mle LL = err ? NA : B.llt
? B.obsy = B.depvar - b0
? B.obsvar = s_obs^2
? B.statevar = s_state^2
? err = kfilter(&B)
? params b0 s_obs s_state
? end mle
Using numerical derivatives
Tolerance = 1.81899e-012
*Reached the maximum iterations (500)*
Error executing script: halting
> end mle
On Tue, 1 Dec 2020 at 17:10, Sven Schreiber <svetosch(a)gmx.net> wrote:
> Am 01.12.2020 um 17:20 schrieb Brian Revell:
>
> Hi Sven
> many thanks for your prompt feedback. Where did you paste it in? In the
> Model MLE option from the GUI -or under File -Script File-menu option?
>
> The latter, yes, so basically in the built-in script editor. The MLE area
> is meant for even lower-level access, without the predefined Kalman filter
> objects.
>
> know where to do this. The manual is not helpful in this respect.
>
> In later parts of the guide it is pretty much assumed that people know
> the scripting interface of gretl. However, at the beginning this is
> explained, please see section 3.1.
>
> (BTW, in my version of the guide the script listing is number 36.4 in ch.
> 36, I guess you have a slightly older version where some newer chapters do
> not exist yet. But this shouldn't make a difference in this context.)
>
> cheers
>
> sven
> _______________________________________________
> Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
> To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
> Website:
> https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
>
10:47 a.m.
Am 17.12.2020 um 20:26 schrieb Brian Revell:
Hi Sven
with the Script below which is for a time varying parameter. there
are 2 things I wish to get from the output of this routine.
First in the final gnuplot command I have added, I want to include in
the graph of the dependent variable the estimated/smoothed level of
the system with its 95% CI
Second to get a printout of the values of tvar_ bhat and its 95% CI
and of the level of the system
Also with regard to StrucTiSM, it would be helpful have a printout of
the values of estimated level of the system (and its 95%CI)
Is there some simple script I can add to do these things.? The
forecast script you supplied works well thank you in StrucTism post
estimation.
Brian, you would have to ask more specific questions, because we cannot
write your scripts for you. We have put your ideas about missing GUI
features in StrucTiSM on our to-do-list, but this is not going to happen
in the very short term. For general handling of matrix and series
objects in gretl (scripting) and the gnuplot integration I can for now
only refer you to the relevant chapters of the user guide. If then you
get an error which you don't understand, you're very welcome to post it
here and ask for help.
sorry,
sven
1:14 p.m.
OK thanks
i'll look forward to seeing the next version of StrucTiSM.
Not everyone needs to programme in R or Hansi nor unpick the code of
packages, their interest bing in using the software to solve real world
problems by applying academically accepted methodologies.
Happy to alert you to any other issues that might arise leading to error
messages. Please add the queries in my message requesting script -it would
be useful to other users I am certain.
cheers
Brian
On Fri, 18 Dec 2020 at 10:47, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 17.12.2020 um 20:26 schrieb Brian Revell:
Hi Sven
with the Script below which is for a time varying parameter. there are 2
things I wish to get from the output of this routine.
First in the final gnuplot command I have added, I want to include in the
graph of the dependent variable the estimated/smoothed level of the
system with its 95% CI
Second to get a printout of the values of tvar_ bhat and its 95% CI and
of the level of the system
Also with regard to StrucTiSM, it would be helpful have a printout of the
values of estimated level of the system (and its 95%CI)
Is there some simple script I can add to do these things.? The forecast
script you supplied works well thank you in StrucTism post estimation.
Brian, you would have to ask more specific questions, because we cannot
write your scripts for you. We have put your ideas about missing GUI
features in StrucTiSM on our to-do-list, but this is not going to happen in
the very short term. For general handling of matrix and series objects in
gretl (scripting) and the gnuplot integration I can for now only refer you
to the relevant chapters of the user guide. If then you get an error which
you don't understand, you're very welcome to post it here and ask for help.
sorry,
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
8:15 p.m.
Am 18.12.2020 um 14:14 schrieb Brian Revell:
OK thanks
i'll look forward to seeing the next version of StrucTiSM.
This version is now published and online. It supports forecasting from
the GUI, and also provides the resulting series as part of the results
that you should be able to add to your dataset.
Update your version from within gretl (File - Function packages - On
server). Let us know if you spot anything which isn't working.
thanks
sven
12:01 p.m.
HI Sven
Happy New Year
Thanks for the info re state space forecasting.
I'll give it a workout you can be assured.
Brian
On Sat, 2 Jan 2021 at 20:15, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 18.12.2020 um 14:14 schrieb Brian Revell:
> OK thanks
> i'll look forward to seeing the next version of StrucTiSM.
This version is now published and online. It supports forecasting from
the GUI, and also provides the resulting series as part of the results
that you should be able to add to your dataset.
Update your version from within gretl (File - Function packages - On
server). Let us know if you spot anything which isn't working.
thanks
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
12:47 p.m.
State-space forecasts
The forecast part of the package works well as far as I can see. However,
it would be helpful and useful if the fitted values of the BSM are printed
out or retrievable , in order to be able to create plots that show original
dep. variable, fitted and forecast values on the same plot. Currently
there one plot which shows dep var and fitted values, and another which
shows dep var and forecast. Apparently never the two shall meet. A print
out of fitted values would also be useful -otherwise it is somewhat tedious
picking them off the plot readout.
I assume the shaded area of the forecast plot is the 95% CI - SEs for the
fitted values would also be a useful information in the estimation results.
Cheers
Brian
On Sat, 2 Jan 2021 at 20:15, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 18.12.2020 um 14:14 schrieb Brian Revell:
> OK thanks
> i'll look forward to seeing the next version of StrucTiSM.
This version is now published and online. It supports forecasting from
the GUI, and also provides the resulting series as part of the results
that you should be able to add to your dataset.
Update your version from within gretl (File - Function packages - On
server). Let us know if you spot anything which isn't working.
thanks
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
4:07 p.m.
Am 03.01.2021 um 13:47 schrieb Brian Revell:
State-space forecasts
The forecast part of the package works well as far as I can see.
Hi, thanks for testing.
However, it would be helpful and useful if the fitted values of the
BSM
are printed out or retrievable , in order to be able to create plots
that show original dep. variable, fitted and forecast values on the same
plot. Currently there one plot which shows dep var and fitted values,
and another which shows dep var and forecast. Apparently never the two
shall meet. A print out of fitted values would also be useful
-otherwise it is somewhat tedious picking them off the plot readout.
What you can do already is to save the series with the suffixes _level,
_irreg, _seas, _slope, and then also the matrix "coeff". If you stuff
those series into a gretl list then you can get the fitted values with
the function call:
series fitted = lincomb(mylist, coeff)
But I admit it's a lot of manual labor in the GUI there. We might add
the y-hats or the u-hats as series to the bundle as well, in the future.
I assume the shaded area of the forecast plot is the 95% CI - SEs
for
the fitted values would also be a useful information in the estimation
results.
Yes, 95% as the title of the plot says.
The series of forecast error variances is available in the result bundle
(and thus by clicking that icon) as "serfcv". Not sure what you mean by
"SE for fitted values".
cheers
sven
1:37 a.m.
As a user rather than programmer, has anyone written script for a state
space forecast using the Kalman filter for gretl?
cheers
Brian
On Tue, 1 Dec 2020 at 15:52, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 01.12.2020 um 15:58 schrieb bjr.newmail(a)gmail.com:
> I wanted to estimate a model with a time varying parameter (on a simple
linear trend) and would appreciate some guidance as I am not a programmer
in Gretl's language.
> 1) Where in the GUI Model set of commands do you access the State Space
modelling - is by writing the spec in the MLE option?
> 2) I loaded the example file AWM.gdt
> 3)Entered the commands below exactly as specified in the manual viz.
Listing 34.4: Phillips curve on Euro data with time-varying slope
> 4) It died at the first line.
>
Hi, I copied and pasted the code from your email and it runs fine here.
Please provide more details about the error message you're getting and
your gretl version etc.
As for your general question: AFAIK there is currently no GUI access to
the general state-space framework yet, given the complexity involved.
For certain special cases the contributed package StrucTiSM offers a GUI
access, among which the local level, the local linear trend, and the
so-called basic structural models. Don't know if it really covers what
you're looking for, though.
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
9:29 a.m.
Am 12.12.20 um 02:37 schrieb Brian Revell:
As a user rather than programmer, has anyone written script for a
state
space forecast using the Kalman filter for gretl?
cheers
Brian
Hi Brian,
you may have a look at the StrucTiSM package for this. Here is the doc:
http://ricardo.ecn.wfu.edu/gretl/cgi-bin/current_fnfiles/unzipped/StrucTi...
Artur
10:51 a.m.
Hi Artur
many thanks.
this is exactly what I need especially as it has a GUI front end.
Is there a URL to download it from?
Brian
On Sat, 12 Dec 2020 at 09:29, Artur Tarassow <atecon(a)posteo.de> wrote:
Am 12.12.20 um 02:37 schrieb Brian Revell:
> As a user rather than programmer, has anyone written script for a state
> space forecast using the Kalman filter for gretl?
> cheers
> Brian
Hi Brian,
you may have a look at the StrucTiSM package for this. Here is the doc:
http://ricardo.ecn.wfu.edu/gretl/cgi-bin/current_fnfiles/unzipped/StrucTi...
Artur
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
10:54 a.m.
Am 12.12.20 um 11:51 schrieb Brian Revell:
Hi Artur
many thanks.
this is exactly what I need especially as it has a GUI front end.
Is there a URL to download it from?
However, currently StrucTiSM only supports forecasting with no exogenous
stochastic regressors. It's on my list to make this more flexible though...
Packages can be downloaded + "installed" either through the GUI or by
running:
pkg install StrucTiSM
To load the package simply run:
include StrucTiSM.gfn
Best,
Artur
10:54 a.m.
On Sat, 12 Dec 2020, Brian Revell wrote:
Hi Artur
many thanks.
this is exactly what I need especially as it has a GUI front end.
Is there a URL to download it from?
As with all official gretl packages, you gust open gretl and go to
Files -> Function Packages -> On server
HTH,
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
4:09 p.m.
Thanks this has been helpful thus far. Next question, having run just the
basic Structural TS model, how do generate the forecasts as the the GUI
only produces the graph of the fitted model vs actual. There is no link to
Analysis. So do I need to save the model as a bundle and feed into another
separate downloadable function to produce the forecast variable plus CI? Is
there yet another function I need to pass the parameters to to see the
smoothed series
cheers
Brian
Brian J Revell
Professor Emeritus
Agricultural and Food Economics
Harper Adams University
Newport
Shropshire
TF10 8NB
Tel 01952815237.
Sent from my Galaxy S8+
On Sat, 12 Dec 2020 at 10:54, Riccardo (Jack) Lucchetti <
p002264(a)staff.univpm.it> wrote:
On Sat, 12 Dec 2020, Brian Revell wrote:
> Hi Artur
> many thanks.
> this is exactly what I need especially as it has a GUI front end.
> Is there a URL to download it from?
As with all official gretl packages, you gust open gretl and go to
Files -> Function Packages -> On server
HTH,
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
4:22 p.m.
Am 12.12.20 um 17:09 schrieb Brian Revell:
Thanks this has been helpful thus far. Next question, having run just
the basic Structural TS model, how do generate the forecasts as the
the GUI only produces the graph of the fitted model vs actual. There
is no link to Analysis. So do I need to save the model as a bundle and
feed into another separate downloadable function to produce the
forecast variable plus CI? Is there yet another function I need to
pass the parameters to to see the smoothed series
Brian did you see the sample script for StrucTiSM? It includes an
example on how to compute the forecasts.
Click File -> Function packages -> On local machine and the then right
click on StrucTiSM -> Sample script.
Best,
Artur
4:37 p.m.
Am 12.12.2020 um 17:22 schrieb Artur Tarassow:
Am 12.12.20 um 17:09 schrieb Brian Revell:
> Thanks this has been helpful thus far. Next question, having run just
> the basic Structural TS model, how do generate the forecasts as the
> the GUI only produces the graph of the fitted model vs actual. There
> is no link to Analysis. So do I need to save the model as a bundle and
> feed into another separate downloadable function to produce the
> forecast variable plus CI? Is there yet another function I need to
> pass the parameters to to see the smoothed series
Brian did you see the sample script for StrucTiSM? It includes an
example on how to compute the forecasts.
Click File -> Function packages -> On local machine and the then right
click on StrucTiSM -> Sample script.
That's a useful pointer, but it is also true that we should probably try
to enable forecasting via the package's output window. (I'm a co-author
of the package.)
(Internal question: the packaging guide says in section 4.3 that a
"bundle-fcast" designation is part of the spec, but that it's "unused
at
present". Does that mean it doesn't work yet?)
In the meantime what you can do is indeed to save the bundle from the
package's output window, say under the name "bstrucTS" or whatever. Then
in a script window or gretl console you'd need to do (without the hansl
tags):
<hansl>
include StrucTiSM.gfn
STSM_fcast( &bstrucTS, 12)
series depvarfcast = bstrucTS.fcast
series depvarfcvar = bstrucTS.fcastvar
</hansl>
Here the number 12 is the arbitrary forecast horizon. You need to make
sure you have those observations in the dataset beyond the estimation
sample, possibly by adding observations before you do the forecast.
BTW, good to have someone testing the package who is able to report any
problems on the list!
thanks
sven
7:05 p.m.
Hi Sven
thanks for your reply. I much appreciate the response rate and assistance
of the staff's prompt support to me as a Gretl user.
I have downloaded the script to look at what underlies the GUI driver. As a
an applied economist and not a software developer (although I was brought
up on econometrics with both Johnstone and Jenkins back in the '70s and
early 80's -yeah that dates me), I have always gone for GUI driven
software when available. Although it may restrict some analytical options
there is usually enough to do modelling the data, model selection and
teasing out the interpretation of the analysis for policy-makers, or even
just intelligent but non-technical readers.
I will look at the graphics script and see what I can make of it adding it
into the appropriate Structural model script. I had saved the model as a
bundle, but when I looked at it couldn't see anything there. I did manage
to use a script file from the manual to estimate a time varying trend, and
time varying parameter, though that required minimum editing to the file.
Again, getting a forecast then became a call for assistance. It would be a
useful and logical addition to the StrucTiSM suite of model options, in the
same way that the Analysis section of modelling output will produce
static or dynamic forecasts from the ARMA modelling option in the Gretl
GUI. Another thought is that it would also be useful in Gretl to be able
to save or print out the TS model in equation form (as is possible for
example in OLS model estimation) -although perhaps that is not so
straightforward to do in multiplying the lag operators.
I'll try your suggestions in the morning -I am sure they should work -it is
more a question of whether I can!
Thanks again for yoour help
Brian
er <svetosch(a)gmx.net> wrote:
Am 12.12.2020 um 17:22 schrieb Artur Tarassow:
> Am 12.12.20 um 17:09 schrieb Brian Revell:
>> Thanks this has been helpful thus far. Next question, having run just
>> the basic Structural TS model, how do generate the forecasts as the
>> the GUI only produces the graph of the fitted model vs actual. There
>> is no link to Analysis. So do I need to save the model as a bundle and
>> feed into another separate downloadable function to produce the
>> forecast variable plus CI? Is there yet another function I need to
>> pass the parameters to to see the smoothed series
>
> Brian did you see the sample script for StrucTiSM? It includes an
> example on how to compute the forecasts.
>
> Click File -> Function packages -> On local machine and the then right
> click on StrucTiSM -> Sample script.
That's a useful pointer, but it is also true that we should probably try
to enable forecasting via the package's output window. (I'm a co-author
of the package.)
(Internal question: the packaging guide says in section 4.3 that a
"bundle-fcast" designation is part of the spec, but that it's "unused
at
present". Does that mean it doesn't work yet?)
In the meantime what you can do is indeed to save the bundle from the
package's output window, say under the name "bstrucTS" or whatever. Then
in a script window or gretl console you'd need to do (without the hansl
tags):
<hansl>
include StrucTiSM.gfn
STSM_fcast( &bstrucTS, 12)
series depvarfcast = bstrucTS.fcast
series depvarfcvar = bstrucTS.fcastvar
</hansl>
Here the number 12 is the arbitrary forecast horizon. You need to make
sure you have those observations in the dataset beyond the estimation
sample, possibly by adding observations before you do the forecast.
BTW, good to have someone testing the package who is able to report any
problems on the list!
thanks
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
5:08 p.m.
Hi Sven, having estimated a local linear trend model via the GUI, it
enables me to save the bundle -but actually all that happensis it lists the
content, but does not enable me to give it a name. If I go to file and use
the command save bundle, all it saves is the estimation output. Also it is
not possible to retreive the fitted values -only a graph.
So I have been unable to see any way to save and link the Structural time
series model outputs to the script you kindly sent (minus the hansi
markers). It would be good to be able to not only produe the forecasts but
the 95% CI around them, and graph in the same way as one can from the
example in the Gretl manual for the time variable parameter.
cheers
Brian
On Sat, 12 Dec 2020 at 16:37, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 12.12.2020 um 17:22 schrieb Artur Tarassow:
> Am 12.12.20 um 17:09 schrieb Brian Revell:
>> Thanks this has been helpful thus far. Next question, having run just
>> the basic Structural TS model, how do generate the forecasts as the
>> the GUI only produces the graph of the fitted model vs actual. There
>> is no link to Analysis. So do I need to save the model as a bundle and
>> feed into another separate downloadable function to produce the
>> forecast variable plus CI? Is there yet another function I need to
>> pass the parameters to to see the smoothed series
>
> Brian did you see the sample script for StrucTiSM? It includes an
> example on how to compute the forecasts.
>
> Click File -> Function packages -> On local machine and the then right
> click on StrucTiSM -> Sample script.
That's a useful pointer, but it is also true that we should probably try
to enable forecasting via the package's output window. (I'm a co-author
of the package.)
(Internal question: the packaging guide says in section 4.3 that a
"bundle-fcast" designation is part of the spec, but that it's "unused
at
present". Does that mean it doesn't work yet?)
In the meantime what you can do is indeed to save the bundle from the
package's output window, say under the name "bstrucTS" or whatever. Then
in a script window or gretl console you'd need to do (without the hansl
tags):
<hansl>
include StrucTiSM.gfn
STSM_fcast( &bstrucTS, 12)
series depvarfcast = bstrucTS.fcast
series depvarfcvar = bstrucTS.fcastvar
</hansl>
Here the number 12 is the arbitrary forecast horizon. You need to make
sure you have those observations in the dataset beyond the estimation
sample, possibly by adding observations before you do the forecast.
BTW, good to have someone testing the package who is able to report any
problems on the list!
thanks
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
6:24 p.m.
Am 15.12.2020 um 18:08 schrieb Brian Revell:
Hi Sven, having estimated a local linear trend model via the GUI, it
enables me to save the bundle -but actually all that happensis it
lists the content, but does not enable me to give it a name. If I go
to file and use the command save bundle, all it saves is the
estimation output. Also it is not possible to retreive the fitted
values -only a graph.
When I execute a specification with StrucTiSM graphically, I get a new
gretl window with the title "Structural TS" and the printout. The
leftmost button has a diskette icon and on clicking I can select the
second menu line "Save bundle to session as icon" or s.th. to that effect.
Can you do that on your installtion as well? If not, please describe
what exactly is absent.
So I have been unable to see any way to save and link the Structural
time series model outputs to the script you kindly sent (minus the
hansi markers). It would be good to be able to not only produe the
forecasts but the 95% CI around them, and graph in the same way as one
can from the example in the Gretl manual for the time variable parameter.
Agreed. We will clarify among the gretl team whether the bundle-fcast
apparatus is already functional and could be used by a package. (We
still would have to update the StrucTiSM package, of course, but
shouldn't be too difficult.)
cheers
sven
7:34 p.m.
Hi Sven
yes I cAn save bundle as an icon. But when I open the bundle icon, all
there is is the output printout, whereas theafolder icon lists a heap of
components from the estimation though no values, just descriptors. However
if one of them is clicked, it reveaks the content if the element, but there
is no obvious way to access them or the folder if a forecast script module
were to be run subsequently.
Brian
Brian J Revell
Professor Emeritus
Agricultural and Food Economics
Harper Adams University
Newport
Shropshire
TF10 8NB
Tel 01952815237.
Sent from my Galaxy S8+
On Tue, 15 Dec 2020, 18:24 Sven Schreiber, <svetosch(a)gmx.net> wrote:
Am 15.12.2020 um 18:08 schrieb Brian Revell:
Hi Sven, having estimated a local linear trend model via the GUI, it
enables me to save the bundle -but actually all that happensis it lists the
content, but does not enable me to give it a name. If I go to file and use
the command save bundle, all it saves is the estimation output. Also it is
not possible to retreive the fitted values -only a graph.
When I execute a specification with StrucTiSM graphically, I get a new
gretl window with the title "Structural TS" and the printout. The leftmost
button has a diskette icon and on clicking I can select the second menu
line "Save bundle to session as icon" or s.th. to that effect.
Can you do that on your installtion as well? If not, please describe what
exactly is absent.
So I have been unable to see any way to save and link the Structural time
series model outputs to the script you kindly sent (minus the hansi
markers). It would be good to be able to not only produe the forecasts but
the 95% CI around them, and graph in the same way as one can from the
example in the Gretl manual for the time variable parameter.
Agreed. We will clarify among the gretl team whether the bundle-fcast
apparatus is already functional and could be used by a package. (We still
would have to update the StrucTiSM package, of course, but shouldn't be too
difficult.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
10:44 p.m.
Am 15.12.2020 um 20:34 schrieb Brian Revell:
Hi Sven
yes I cAn save bundle as an icon. But when I open the bundle icon, all
there is is the output printout, whereas theafolder icon lists a heap of
components from the estimation though no values, just descriptors.
However if one of them is clicked, it reveaks the content if the
element, but there is no obvious way to access them or the folder if a
forecast script module were to be run subsequently.
You shouldn't need to "open" the bundle or access it through the GUI.
Once it's in the session it should be there and ready for the scripting
code to access it. Have you tried running the code lines I posted
before? (With the name suitably adapted.)
cheers
sven
9:13 a.m.
Hi Sven
yes I have tried it -without success
The bundle was saved after estimation using the Save File command as
SevernSTS . The Bundle icon is attached as a jpg. I also saved the bundle
contents but that command does not enable me to name a file so I don't know
if that info was actually saved. it doesn't appear in the icon attached.
I then went to the Gretl file command and created a script file containing
the text you sent to me having changed the name of the download bundle file
name and dependent variable -it clearly couldn't find/include the Bundle
file I had saved. I have attached an image of the Bundle SevernSTS. As I
said, however, this only appears to contain the estimation output file not
the individual elements of the estimation process which are listed in the
Command Save Bundle content .
Script file run:-
----------------------------------------------------------
include SevernSTS.gfn
STSM_fcast( &bstrucTS, 5)
series SEVERN_EGGS_UNADJ = bstrucTS.fcast
series SEVERN_EGGS_UNADJ = bstrucTS.fcastvar
-----------------------------------------------
This was the log of running the script file:-
gretl version 2020e
Current session: 2020-12-16 08:50
? include SevernSTS.gfn
Error reading SevernSTS.gfn
Error executing script: halting
include SevernSTS.gfn
----------------------------------------
Cheers
Brian
On Tue, 15 Dec 2020 at 22:44, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 15.12.2020 um 20:34 schrieb Brian Revell:
> Hi Sven
> yes I cAn save bundle as an icon. But when I open the bundle icon, all
> there is is the output printout, whereas theafolder icon lists a heap of
> components from the estimation though no values, just descriptors.
> However if one of them is clicked, it reveaks the content if the
> element, but there is no obvious way to access them or the folder if a
> forecast script module were to be run subsequently.
You shouldn't need to "open" the bundle or access it through the GUI.
Once it's in the session it should be there and ready for the scripting
code to access it. Have you tried running the code lines I posted
before? (With the name suitably adapted.)
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
1 p.m.
Am 16.12.2020 um 10:13 schrieb Brian Revell:
Hi Sven
yes I have tried it -without success
The bundle was saved after estimation using the Save File command as
SevernSTS . The Bundle icon is attached as a jpg. I also saved the
bundle contents but that command does not enable me to name a file so I
don't know if that info was actually saved. it doesn't appear in the
icon attached.
I then went to the Gretl file command and created a script file
containing the text you sent to me having changed the name of the
download bundle file name and dependent variable -it clearly couldn't
find/include the Bundle file I had saved. I have attached an image of
the Bundle SevernSTS. As I said, however, this only appears to contain
the estimation output file not the individual elements of the estimation
process which are listed in the Command Save Bundle content .
Script file run:-
----------------------------------------------------------
include SevernSTS.gfn
The "include" refers to the package name, so it has to be StrucTiSM.gfn.
(This is necessary to make the package known to the script environment,
even though you ran the package before through the GUI.)
STSM_fcast( &bstrucTS, 5)
And judging from your screenshot the bundle name is SevernSTS, so this
call has to be STSM_fcast( &SevernSTS, 5).
series SEVERN_EGGS_UNADJ = bstrucTS.fcast
series SEVERN_EGGS_UNADJ = bstrucTS.fcastvar
same here
cheers
sven
2:48 p.m.
Thanks Sven
progress being made...
A couple of error messages below the 5 period forecast data though they
didn't seem to be terminal.
-------------------------------------------------------------------------------------------------------
gretl version 2020e
Current session: 2020-12-16 13:55
? include StrucTiSM.gfn
C:\Users\Brian\AppData\Roaming\gretl\functions\StrucTiSM\StrucTiSM.gfn
? STSM_fcast(&SevernSTS,5)
Out of sample forecast for SEVERN_EGGS_UNADJ
horizon forecast std.err.
1 7.9429 4.85375
2 7.85103 5.71423
3 7.75916 6.51948
4 7.6673 7.28785
5 7.57543 8.02992
? series SEVERN_EGGS_UNADJ = SevernSTS.fcast
Incompatible types in assignment: series = matrix
Error executing script: halting
series SEVERN_EGGS_UNADJ = SevernSTS.fcast
----------------------------------------------------------------------
My next query is how I can get the values of the model system state that
are shown together with the actual data series via the graphics tab .
Ideally I would like to show these with their 95% CI, and also the
contiguous forecast values with their 95% CI . The second output that
would be useful would be a smoothed value of the time varying trend
co-efficient from the model. Do the values of the time variable default
to 1,2.....n or do they take the date values in the estimation.?
Thanks again for your guidance .
best wishes
Brian
On Wed, 16 Dec 2020 at 13:00, Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 16.12.2020 um 10:13 schrieb Brian Revell:
> Hi Sven
> yes I have tried it -without success
>
> The bundle was saved after estimation using the Save File command as
> SevernSTS . The Bundle icon is attached as a jpg. I also saved the
> bundle contents but that command does not enable me to name a file so I
> don't know if that info was actually saved. it doesn't appear in the
> icon attached.
> I then went to the Gretl file command and created a script file
> containing the text you sent to me having changed the name of the
> download bundle file name and dependent variable -it clearly couldn't
> find/include the Bundle file I had saved. I have attached an image of
> the Bundle SevernSTS. As I said, however, this only appears to contain
> the estimation output file not the individual elements of the estimation
> process which are listed in the Command Save Bundle content .
> Script file run:-
> ----------------------------------------------------------
> include SevernSTS.gfn
The "include" refers to the package name, so it has to be StrucTiSM.gfn.
(This is necessary to make the package known to the script environment,
even though you ran the package before through the GUI.)
> STSM_fcast( &bstrucTS, 5)
And judging from your screenshot the bundle name is SevernSTS, so this
call has to be STSM_fcast( &SevernSTS, 5).
> series SEVERN_EGGS_UNADJ = bstrucTS.fcast
> series SEVERN_EGGS_UNADJ = bstrucTS.fcastvar
same here
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
2:53 p.m.
Am 16.12.2020 um 15:48 schrieb Brian Revell:
Thanks Sven
progress being made...
A couple of error messages below the 5 period forecast data though they
didn't seem to be terminal.
-------------------------------------------------------------------------------------------------------
gretl version 2020e
Current session: 2020-12-16 13:55
? include StrucTiSM.gfn
C:\Users\Brian\AppData\Roaming\gretl\functions\StrucTiSM\StrucTiSM.gfn
? STSM_fcast(&SevernSTS,5)
Out of sample forecast for SEVERN_EGGS_UNADJ
horizon forecast std.err.
1 7.9429 4.85375
2 7.85103 5.71423
3 7.75916 6.51948
4 7.6673 7.28785
5 7.57543 8.02992
? series SEVERN_EGGS_UNADJ = SevernSTS.fcast
Incompatible types in assignment: series = matrix
The background here is that the active sample doesn't match the length
of the matrix. (rows)
This could actually be a bug in our package, since the match should be
checked I guess. Will have a look later.
cheers
sven
5:02 p.m.
Hi Sven
i HAVE ALSO BEEN LOOKING AT THE VARIABEL SLOPE COEFFICENT MODEL IN THE USER
GUIDE IN WHICH THE EXOGENOUS VARIABLE IS TIME TIME IE IT WILL BE A VARIABLE
TREND COEFFICIENT
THIS IS MY SCRIPT ADAPTED FROM THE USER MANUAL WHICH WORKS WELL
BUT I ALSO WANT TO GRAPH THE ACTUAL DATA, TOGETHER WITH THE SMOOTHED
STATE VALUES (AS INDICATED IN THE FINAL LINE) AND TO BE ABLE TO PRINT THE
SMOOTHED SLOPE AND STATE VALUES OF THE DEP VAR.
function void at_each_step(bundle *b)
b.obsymat = transp(b.mX[b.t,])
end function
open SevernDeeEggs.gdt --quiet
smpl 1998 2019
/* parameter initialization */
scalar b0 = mean(SEVERN_EGGS_UNADJ)
scalar s_obs = 0.1
scalar s_state = 0.1
/* bundle setup */
bundle B = ksetup(SEVERN_EGGS_UNADJ, 1, 1, 1)
matrix B.mX = {time}
matrix B.depvar = {SEVERN_EGGS_UNADJ}
B.timevar_call = "at_each_step"
B.diffuse = 1
set bfgs_maxiter 5000
/* ML estimation of intercept and the two variances */
mle LL = err ? NA : B.llt
B.obsy = B.depvar - b0
B.obsvar = s_obs^2
B.statevar = s_state^2
err = kfilter(&B)
params b0 s_obs s_state
end mle
/* display the smoothed time-varying slope */
ksmooth(&B)
series tvar_b1hat = B.state[,1]
series tvar_b1se = sqrt(B.stvar[,1])
gnuplot tvar_b1hat --time-series --with-lines --output=display \
--band=tvar_b1hat,tvar_b1se,1.96 --band-style=fill
gnuplot SEVERN_EGGS_UNADJ *K SMOOTHED STATE VAR* --time-series --with-lines
--output=display
Any suggestions welcomed as what script I need to add to produce a printout
of the estimated values .
cheers
Brian
2:58 p.m.
Am 16.12.2020 um 15:53 schrieb Sven Schreiber:
Am 16.12.2020 um 15:48 schrieb Brian Revell:
>
> ? series SEVERN_EGGS_UNADJ = SevernSTS.fcast
> Incompatible types in assignment: series = matrix
The background here is that the active sample doesn't match the length
of the matrix. (rows)
This could actually be a bug in our package, since the match should be
checked I guess. Will have a look later.
Sorry, that was premature - the command line above is yours I think, not
internally inside the package. So it's not our bug. To stuff a vector
(matrix type) of, say, 5 elements into a gretl series that shows up
directly in the dataset, you have to define a matching active sample of
5 obs. For a forecast chances are that first you have to extend the
dataset range by adding new obs. All that can be done from the menus.
Then try again with the command line above.
good luck
sven
7:45 p.m.
Evening Artur
I shall be looking at it tomorrow thanks -after decorating a Christmas tree!
cheers
Brian
On Sat, 12 Dec 2020 at 16:22, Artur Tarassow <atecon(a)posteo.de> wrote:
Am 12.12.20 um 17:09 schrieb Brian Revell:
> Thanks this has been helpful thus far. Next question, having run just
> the basic Structural TS model, how do generate the forecasts as the
> the GUI only produces the graph of the fitted model vs actual. There
> is no link to Analysis. So do I need to save the model as a bundle and
> feed into another separate downloadable function to produce the
> forecast variable plus CI? Is there yet another function I need to
> pass the parameters to to see the smoothed series
Brian did you see the sample script for StrucTiSM? It includes an
example on how to compute the forecasts.
Click File -> Function packages -> On local machine and the then right
click on StrucTiSM -> Sample script.
Best,
Artur
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
1449
days inactive
1482
days old
35 comments
5 participants
participants (5)
-
Artur Tarassow -
bjr.newmail＠gmail.com -
Brian Revell
-
Riccardo (Jack) Lucchetti -
Sven Schreiber