selection of ARIMA model
by Sam Sam
Dear all:
step one:
I use a set of data and select two appropriate ARIMA model (parameter is significant, and residual is white noise), they are ARIMA(1,1,0) and ARIMA(3,1,2).(their observations are different because lags of AR or MA are different).
step two:
I want to choose one from them by AIC and BIC. But there comes a question. In order to compare their AIC and BIC, I have to set the sample range of these two models the same (their observations are the same now). But the new ARIMA(1,1,0) and original ARIMA(1,1,0) are different models (for example, their parameters are different. Residual of original model is white noise but new model is not). The new ARIMA(3,1,2) and original ARIMA(3,1,2) are also different models. If AIC and BIC of new ARIMA(1,1,0) is lower than AIC and BIC of new ARIMA(3,1,2), which model should I select? New ARIMA(1,1,0) or original ARIMA(1,1,0)
p.s. Is there any wrong on the procedure of selection of ARIMA model? Should I set the sample range the same at start (step one)?
Thanks a lot.
10 years, 9 months
microecometric functional form test
by Julio Acuña
I have a question regarding functional forma when working with complex surveys.
I'm trying to choose between linear and other funciontal forms and I'd like to know
if it is possible to use Ramsey Reset test, makinnon's tests , etc . Or it the inclusion of weights and clustered robust estimations limits its use/?
Thanks in advance!
10 years, 9 months
modtest with vars & vecms
by Summers, Peter
Hi all,
I've discovered what I think is a minor bug in the modtest command: if I use it right after estimating a VAR or VECM, the --quiet option has no effect. The auxilliary regressions are printed with or without that option.
I also have a question about the $test and $pvalue accessors in these cases. At the moment, they're scalars containing the respective quantities from the test on the last equation. Would it be possible to have them return vectors containing the p-values/test stats from each equation?
TIA,
PS
10 years, 9 months
genr, vs scalar and series
by Lee Adkins
Hi all,
What is the preferred use of these in hansl? I've started using scalar for
all scalar computations (but in any old code I used genr); sometimes I use
series instead of genr to generate new variables. I do realize that series
will generate a series from a scalar.
For instance, which is the preferred usage:
open engel
set seed 3213789
loop 100 --progressive --quiet
series u = normal(0,88)
series y1= 80+10*income+u
ols y1 const income
endloop
or
open engel
set seed 3213789
loop 100 --progressive --quiet
genr u = normal(0,88)
genr y1= 80+10*income+u
ols y1 const income
endloop
also suppose I want to print or store a scalar, which of these is preferred
way to generate the scalar?
scalar b = $coeff(income)
genr b = $coeff(income)
I'm trying to finish the updated version of my gretl book and my code needs
some tidying up. Looking at it now, it is pretty much a mess. It seems
I've mixed the usage of these up throughout and I'd like to know the
preferred way before I hack my way back through a thousand lines of code
.....
Thanks,
Lee
--
Lee Adkins
Professor of Economics
lee.adkins(a)okstate.edu
learneconometrics.com
10 years, 9 months
IV Panel Estimation
by Eloy Arturo Fisher Hogan
Dear Gretl friends,
I've been working to get familiar with a macro panel I'm doing - it comprises a number of variables of US States since 1992 until 2009. However, as everything in econometrics, I have to be very careful with some endogeneity issues. For this reason, I've working on finding suitable IV's for my estimations. However, I have some questions regarding Gretl's panel toolkit.
The first is about the Panel IV GUI and how it relates to the tsls function. I am pretty sure it is improper, yet I will ask nonetheless -- is it still feasible to use tsls for the panel estimation? I am pretty sure you can't, since it doesn't demean the data and the like, but I want to be sure since I am more familiar with the test outputs reported there. If not, how can you script what Allin programmed in the Panel IV GUI? I've not been able to find the script functions for what it is done there (I see there is an ivpanel function in the GUI example but nothing else is said about it). Is it equivalent to using a dpanel with instruments?
Thanks everyone!
EF
10 years, 9 months
AIC and BIC
by Sam Sam
Monthly data for in-sample forecasting : sample from 1980M1 to 1990M12(the number of samples is 132)
Estimating the model:
The using observations of ARIMA(1,1,1) : 1980M3 to 1990M12. (number:130)
The using observations of ARIMA(2,1,0): 1980M4 to 1990M12. (number:129)
The using observations of ARIMA(1,1,0)(1,1,0)12: 1982M3 to 1990M12. (number:106)
Can I compare these three models by AIC or BIC?
In order to make the number of observations of these models the same for comparing AIC and BIC, I have to set the sample range from 1982M3 to 1990M12.
Estimating the model:
The using observations of ARIMA(1,1,1) : 1982M3 to 1990M12. (number:106)
The using observations of ARIMA(2,1,0): 1982M3 to 1990M12. (number:106)
The using observations of ARIMA(1,1,0)(1,1,0)12: 1982M3 to 1990M12. (number:106)
But now the period of in-sample forecasting is not 1980M1 - 1990M12.
The outcome of two ARIMA(1,1,0)(1,1,0)12 are the same.
The outcome of two ARIMA(1,1,1) are different.
The outcome of two ARIMA(2,1,0) are different.
My question is how to choose ARIMA model under in-sample forecasting period is 1980M1 -1990M12 . I want to use sample from 1980M1 to 1990M12 for in-sample forecasting. But in order to compare AIC and BIC, I have to set the sample range from 1982M3 to 1990M12. The outcome of ARIMA(1,1,1) and ARIMA(2,1,0) are different. How do I use the same data to choose ARIMA model from different order of AR and MA (SAR and SMA)?
Can I compare AIC and BIC under in-sample forecasting period is 1980M1 -1990M12 ?
10 years, 9 months
AIC and BIC
by Sam Sam
Dear all:
I want to estimate model using data: 1980M1 to 1990M12
Model estimation:
AR(1) (observation:1980M2-1990M12)
AR(2) (observation:1980M3-1990M12)
AR(3) (observation:1980M4-1990M12)
Can I compare AIC of these three models ?
In order to make the number of observations the same, I have to set the data range from 1980M4 to 1990M12.
Model estimation:
AR(1) (observation:1980M4-1990M12)
AR(2) (observation:1980M4-1990M12)
AR(3) (observation:1980M4-1990M12)
Can I compare AIC of these three models now ?
I am confused. The outcome of AR(1) (also AR(2) ) are different. My original idea is to build the model to fit the data from 1980M1 to 1990M12. But in order to compare AIC, I have to set the data range from 1980M4 to 1990M12. It is away from my original idea. How can I build model for data from 1980M1 to 1990M12 and also compare the AIC from different model?
10 years, 9 months
(no subject)
by Sam Sam
The monthly data from 1980M1 to 1990M12(the number of samples is 132).
When estimating the model,
The using observations of ARIMA(1,1,1) : 1980M3 to 1990M12. (number:130)
The using observations of ARIMA(2,1,0):1980M4 to 1990M12. (number:129)
The using observations of ARIMA(1,1,0)(1,1,0)12:1983M3 to 1990M12. (number:106)
Can I compare AIC or BIC of these three models when the number of using observations are different ?
In the equation of BIC, number of observations is included.
In the equation of AIC, number of observations is not included.
When the number of using observations are different, Can I still compare AIC or BIC of these three models?
How do I choose model by AIC or BIC in gretl?
10 years, 10 months
AIC and BIC
by Sam Sam
The monthly data from 1980M1 to 1990M12(the number of samples is 132).
When estimating the model,
The using observations of ARIMA(1,1,1) : 1980M3 to 1990M12. (number:130)
The using observations of ARIMA(2,1,0):1980M4 to 1990M12. (number:129)
The using observations of ARIMA(1,1,0)(1,1,0)12:1983M3 to 1990M12. (number:106)
The number of samples is 132. But the number of the using observations depends on order of AR and MA(also SAR and SMA).
How can I make them the same for comparing AIC and BIC ?
10 years, 10 months
Ralph M Rodriguez/PO/KAIPERM is out of the office.
by Ralph.M.Rodriguez＠kp.org
I will be out of the office starting 07/25/2011 and will not return until
08/01/2011.
Hi All, I will be out of the office on business from July 25 to July 29
returning to the office Aug 1, 2011 I will be cheking emails in the
evening. But, if you have an immediate concern with Cost Model, email
CM-Help(a)kp.org or call Bradley Njus at 510 625 4595.
Thanks much,
Ralph
10 years, 10 months
