4:49 p.m.
Hi,
I'm new to gretl, so forgive my ignorance. I'm aware that gretl provides
forecasting functionality, but I'm interested in using any of the
time-series models in gretl (AR, ARIMA, GARCH, etc) to interpolate a
section of unknown data. If I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
Then it is easy in gretl to predict the following years (1954, 1955, and
so forth). But if I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
1956 190
1957 185
1958 170
1959 150
I then want to "predict" (or technically interpolate) the values for the
years 1954 and 1955 (the stock value will probably be something between
195 and 190). Is this possible with gretl? If so, how?
Peter
2:07 p.m.
The estimated equation is
Model 1: OLS, using observations 1950-1957 (T = 8)
Dependent variable: Stock
coefficient std. error t-ratio p-value
--------------------------------------------------------
const 225.000 6.21177 36.22 2.95e-08 ***
Time -6.25000 0.976086 -6.403 0.0007 ***
Mean dependent var 190.6250 S.D. dependent var 22.90313
Sum squared resid 468.7500 S.E. of regression 8.838835
R-squared 0.872340 Adjusted R-squared 0.851064
F(1, 6) 41.00000 P-value(F) 0.000684
Log-likelihood -27.63402 Akaike criterion 59.26804
Schwarz criterion 59.42692 Hannan-Quinn 58.19644
rho -0.010000 Durbin-Watson 1.650000
Now substitute time values in the estimated equation
225- 6.25 X
Give X values like 1 , 2 3 up to 10
You will get predicted values like this ( by manual calculation)
when X is 5 ( value for 1954 ) the estimated value is 193.75. When X is 6 (
value for 1955) the predicted value is 187.5. By OLS this is the answer.
On Sat, May 17, 2014 at 7:19 PM, Huffelpuff <huffelpuff420(a)gmail.com> wrote:
Hi,
I'm new to gretl, so forgive my ignorance. I'm aware that gretl provides
forecasting functionality, but I'm interested in using any of the
time-series models in gretl (AR, ARIMA, GARCH, etc) to interpolate a
section of unknown data. If I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
Then it is easy in gretl to predict the following years (1954, 1955, and
so forth). But if I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
1956 190
1957 185
1958 170
1959 150
I then want to "predict" (or technically interpolate) the values for the
years 1954 and 1955 (the stock value will probably be something between
195 and 190). Is this possible with gretl? If so, how?
Peter
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4:13 p.m.
Thanks, this seems to work.
One last question: I was not able to get these values if I only had my
time variable as regressor. I had to add "const" as a regressor too, in
order to get the values (225 and 6.25). What exactly is const and what
values does it contain (I'm unable to edit this variable)?
Thanks
Peter
On 2014/05/18 01:07 PM, Narandra Dashora wrote:
The estimated equation is
Model 1: OLS, using observations 1950-1957 (T = 8)
Dependent variable: Stock
coefficient std. error t-ratio p-value
--------------------------------------------------------
const 225.000 6.21177 36.22 2.95e-08 ***
Time -6.25000 0.976086 -6.403 0.0007 ***
Mean dependent var 190.6250 S.D. dependent var 22.90313
Sum squared resid 468.7500 S.E. of regression 8.838835
R-squared 0.872340 Adjusted R-squared 0.851064
F(1, 6) 41.00000 P-value(F) 0.000684
Log-likelihood -27.63402 Akaike criterion 59.26804
Schwarz criterion 59.42692 Hannan-Quinn 58.19644
rho -0.010000 Durbin-Watson 1.650000
Now substitute time values in the estimated equation
225- 6.25 X
Give X values like 1 , 2 3 up to 10
You will get predicted values like this ( by manual calculation)
when X is 5 ( value for 1954 ) the estimated value is 193.75. When X
is 6 ( value for 1955) the predicted value is 187.5. By OLS this is
the answer.
On Sat, May 17, 2014 at 7:19 PM, Huffelpuff <huffelpuff420(a)gmail.com
<mailto:huffelpuff420@gmail.com>> wrote:
Hi,
I'm new to gretl, so forgive my ignorance. I'm aware that gretl
provides
forecasting functionality, but I'm interested in using any of the
time-series models in gretl (AR, ARIMA, GARCH, etc) to interpolate a
section of unknown data. If I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
Then it is easy in gretl to predict the following years (1954,
1955, and
so forth). But if I have something like this:
Year Stock value
1950 215
1951 220
1952 200
1953 195
1956 190
1957 185
1958 170
1959 150
I then want to "predict" (or technically interpolate) the values
for the
years 1954 and 1955 (the stock value will probably be something
between
195 and 190). Is this possible with gretl? If so, how?
Peter
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu <mailto:Gretl-users@lists.wfu.edu>
http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4:52 p.m.
Am 18.05.2014 15:13, schrieb Huffelpuff:
Thanks, this seems to work.
One last question: I was not able to get these values if I only had my
time variable as regressor. I had to add "const" as a regressor too, in
order to get the values (225 and 6.25). What exactly is const and what
values does it contain (I'm unable to edit this variable)?
"const" is IMHO pretty self-explanatory, namely a constant term, as in
basically all regression packages. Have you ever run a regression before?
Apart from that, you can arbitrarily use any value you like for
"interpolating" your series, if you don't specify any kind of objective
or loss function. Narandra's solution is also arbitrary (which doesn't
mean it's bad).
cheers,
sven
6 p.m.
When the explanatory ( or independent ) variable is zero then the value
taken by explained ( dependent ) variable is the value of the constant .
But if you see any book on statistics the fitted regression is Y= a+ bX. (
It is also written as (Y= mX+c) . So it is a value. . Sven Schreiber is
correct . Mine is a simple way . If you plot that actual and fitted value
it lies at a point where X is zero. Incidentally it may be positive or
negative also .
Feel free to communicate
On Sun, May 18, 2014 at 6:43 PM, Huffelpuff <huffelpuff420(a)gmail.com> wrote:
Thanks, this seems to work.
One last question: I was not able to get these values if I only had my
time variable as regressor. I had to add "const" as a regressor too, in
order to get the values (225 and 6.25). What exactly is const and what
values does it contain (I'm unable to edit this variable)?
Thanks
Peter
On 2014/05/18 01:07 PM, Narandra Dashora wrote:
The estimated equation is
Model 1: OLS, using observations 1950-1957 (T = 8)
Dependent variable: Stock
coefficient std. error t-ratio p-value
--------------------------------------------------------
const 225.000 6.21177 36.22 2.95e-08 ***
Time -6.25000 0.976086 -6.403 0.0007 ***
Mean dependent var 190.6250 S.D. dependent var 22.90313
Sum squared resid 468.7500 S.E. of regression 8.838835
R-squared 0.872340 Adjusted R-squared 0.851064
F(1, 6) 41.00000 P-value(F) 0.000684
Log-likelihood -27.63402 Akaike criterion 59.26804
Schwarz criterion 59.42692 Hannan-Quinn 58.19644
rho -0.010000 Durbin-Watson 1.650000
Now substitute time values in the estimated equation
225- 6.25 X
Give X values like 1 , 2 3 up to 10
You will get predicted values like this ( by manual calculation)
when X is 5 ( value for 1954 ) the estimated value is 193.75. When X is 6
( value for 1955) the predicted value is 187.5. By OLS this is the answer.
On Sat, May 17, 2014 at 7:19 PM, Huffelpuff <huffelpuff420(a)gmail.com>wrote:
> Hi,
>
> I'm new to gretl, so forgive my ignorance. I'm aware that gretl provides
> forecasting functionality, but I'm interested in using any of the
> time-series models in gretl (AR, ARIMA, GARCH, etc) to interpolate a
> section of unknown data. If I have something like this:
>
> Year Stock value
> 1950 215
> 1951 220
> 1952 200
> 1953 195
>
> Then it is easy in gretl to predict the following years (1954, 1955, and
> so forth). But if I have something like this:
>
> Year Stock value
> 1950 215
> 1951 220
> 1952 200
> 1953 195
> 1956 190
> 1957 185
> 1958 170
> 1959 150
>
> I then want to "predict" (or technically interpolate) the values for the
> years 1954 and 1955 (the stock value will probably be something between
> 195 and 190). Is this possible with gretl? If so, how?
>
> Peter
> _______________________________________________
> Gretl-users mailing list
> Gretl-users(a)lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users
>
_______________________________________________
Gretl-users mailing
listGretl-users@lists.wfu.eduhttp://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4012
days inactive
4013
days old
4 comments
3 participants
participants (3)
-
Huffelpuff -
Narandra Dashora -
Sven Schreiber