December 2014 - Gretl-users - gretlml.univpm.it

Re: [Gretl-users] sequential system restrictions

by Michael Boldin

Since VARs can be estimated as independent equations, omit -auto could be used. Or you could apply a OLS stepwise selection method to each equation, or use AIC or BIC to select the best specification from a set of candidates. Maybe someone has a script to perform the steps, but I doubt anyone has set this up to be as simple as --auto.for a system of equations. One reason you may not find anything as simple as an --auto option for VARs is that data mining of this type (setting zero restrictions) is counter to the VAR modelling philosophy. Also it is rarely necessary because setting lag coefficients to exactly zero when the OLS estimates are statistically similar to zero will typically have few statistically discernible effects in the performance of the entire system (that's why they are statistically insignificant). GRETL does have a VAR lag selection tool that does not violate the general VAR model building concept, but this is probably different from what you desire. Lasso in R might be the most reasonable alternative, especially if you are using the VAR to forecast. Or consider Bayesian VAR modeling that lets you shrink the lagged coefficients toward around priors that are centered around zero. Check out Nobel prize winner Chris Sims R code to implement ( http://www.princeton.edu/~sims/#VARtools)

10 years, 10 months

1
0
0 / 0

sequential system restrictions

by Artur T.

Dear all, I was looking for tools for imposing zero restrictions into a the coefficient matrix of a VAR model. I know that JMulti offers some algorithms to conduct some kind of data mining. For R I found "fastVAR" which uses Lasso penalty to return sparse coefficient matrices. I just would like to ask if anybody in this list has already done some work in this direction using gretl. Actually, I am looking for a very simple algorithm which is comparable to the "omit --auto" command in gretl for single equations. Best wishes, Artur

10 years, 10 months

1
0
0 / 0

How do i get the pvalues from a cointegration test (coint2)?

by Juan Ceccarelli Arias

Hi I'm performing a simple loop like this *** loop i=10...15 coint2 i var1 var2 endloop *** The regular output is like this (i=12) : Contraste de Johansen: Número de ecuaciones = 2 Orden del retardo = 12 Periodo de estimación: 2004:02 - 2014:09 (T = 128) Caso 5: Tendencia y constante no restringidas Log-verosimilitud = 841,346 (Incluyendo un término constante: 478,098) Rango Valor propio Estad. traza Valor p Estad. Lmáx Valor p 0 0,24330 43,795 [0,0000] 35,685 [0,0000] 1 0,061398 8,1105 [0,0044] 8,1105 [0,0044] Corregido por el tamaño muestral (gl = 91) Rango Estad. traza Valor p 0 43,795 [0,0000] 1 8,1105 [0,0050] valor propio 0,24330 0,061398 beta (vectores cointegrantes) varx -2,4516 -0,63655 vary -97,692 32,601 alfa (vectores de ajuste) varx 0,058179 0,026932 vary 0,0035638 -0,0018341 beta renormalizado varx 1,0000 -0,019525 vary 39,849 1,0000 alfa renormalizado varx -0,14263 0,87802 vary -0,0087368 -0,059792 matriz de largo plazo (alfa * beta') varx vary varx -0,15977 -4,8056 vary -0,0075693 -0,40795 ---------------------------------------------------- But i only need the pvalues associated. I mean p Value [0,0000] [0,0044] p Value [0,0000] [0,0050] Under stata i know that if i type return list i would get those values. Well, that's all. Thanks for your time and interest.

10 years, 10 months

2
1
0 / 0

Monte Carlo simulation

by Stefano Fachin

dear Alessandra, although this list is not meant to do students' assignments, your sincere admission deserves a friendly hand :-) On the spot I can see two problems * you must always fix the seed of the random number generator to make sure you are always using the same numbers (for instance, set seed 1234).If this is not clear have a look at the way random numbers are generated, even Wikipedia will do. * to appreciate the effect of increasing sample size you must compare estimates computed from N observations and from a larger sample including those observations plus some more. This is more easily done the other way round: create a dataset, compute your stuff, throw away some observations, compute same stuff, compare. good luck! Stefano -- _________________________________________________________________________ Stefano Fachin Professore Ordinario di Statistica Economica Dip. di Scienze Statistiche Università di Roma "La Sapienza" P.le A. Moro 5 - 00185 Roma - Italia Tel. +39-06-49910834 fax +39-06-49910072 web http://stefanofachin.site.uniroma1.it/

10 years, 10 months

1
0
0 / 0

Monte Carlo Simulation

by Alessandra .Cruanes Aguilar

Dear all, I have been having some troubles computing a Monte Carlo simulation. I am a student, so please don't be too horrified by my short experience :) The problem is the following: I am requested to perform a MC simulation on a lagged-dependent variable model with autocorrelated residuals. The estimator for the lagged-dependent variable, the only regressor, should be inconsistent and biased. My concern is that decreasing the sample size the bias decreases, and I guess that that's just wrong. (I also tried using robust std errors, and nothing changes!) here's the code I used: nulldata 200 > setobs --time-series > > scalar n=$nobs > > scalar beta1=0.8 > scalar rho=0.4 > > smpl 1 1 > series u=0 > series y=0 > > smpl 2 200 > series e=randgen(N,0,1) > series u=rho*u(-1)+e > series y=beta1*y(-1)+u > > > ols y y(-1) --simple > > > loop 5000 --progressive --quiet > > > series e=randgen(N,0,1) > series u=rho*u(-1)+e > series y=beta1*y(-1)+u > > ols y y(-1) > > scalar beta1hat= $coeff(y_1) > scalar s1= $stderr(y_1) > scalar bias=$coeff(y_1)-beta1 > genr LB = beta1hat - critical(t,$df,0.025)*s1 > genr UB = beta1hat + critical(t,$df,0.025)*s1 > print beta1 beta1hat bias LB UB > > endloop > > > smpl 50 --random --replace > > loop 5000 --progressive --quiet > > series e=randgen(N,0,1) > series u=rho*u(-1)+e > series y=beta1*y(-1)+u > > ols y y(-1) > > scalar beta1hat= $coeff(y_1) > scalar s1= $stderr(y_1) > scalar bias=$coeff(y_1)-beta1 > genr LB = beta1hat - critical(t,$df,0.025)*s1 > genr UB = beta1hat + critical(t,$df,0.025)*s1 > print beta1 beta1hat bias LB UB > > endloop > Thank you so much

10 years, 10 months

2
1
0 / 0

progress indicator for time-consuming scripts

by Allin Cottrell

This is in response to Peter Summers' observation in http://lists.wfu.edu/pipermail/gretl-users/2014-December/010481.html It's a long story but I'll try to be brief. Prior to gretl 1.9.91 we had a mechanism that appeared in the menubar of a script editor window to indicate progress when a script took appreciable time to complete: a "spinner" plus a "stop" button that could be used to halt execution of the script. (At that time the output window didn't appear until execution of the script was finished.) More recently, I figured it was preferable to open a script output window right away, and show the spinner + stop-button in that window. In this way, we could arrange for partial output to appear as it was produced, giving a more convincing impression that progress was being made (see the newish "flush" command). That worked up to a point, but after Peter's comment I realized that the policy "open an output window right away and show a spinner and stop-button there" was not always engaged; it depended, in a way that I hadn't intended, on certain user choices. I think this is now fixed in CVS and snapshots. Testing would be welcome. If you run a time-consuming script in the GUI you should see the output window "come to the front" with spinner and stopper. (But if the script completes quickly you probably won't notice this.) This is quite tricky to code, and I'd be particularly interested if anyone is able to provoke a crash (e.g. by pressing the stop button). Allin Cottrell

10 years, 10 months

3
2
0 / 0

fortnightly data

by JOSE FRANCISCO PERLES RIBES

Dear list: I know that this is a very basic question. But how can I introduce fortnightly data in Gretl?. I have a series hotel occupancy delivered each two weeks and I want to analyze it. I have used gretl GUI with the non-standard frequencies, but my result is always non-sense. If I use 24 frequencies (12 months x 2 pairs of weeks) Gretl interpret them as 'hourly' data. Thanks in advance and sorry for any inconvenience. José Perles University of Alicante

10 years, 10 months

2
1
0 / 0

Question about join

by Logan Kelly

I am using the join command and getting the following error '2013q4' does not match the format '%Yq%q' But I cannot figure out why the date does not match the format string. I am probably just making a silly mistake. Thanks, Logan -- Logan J. Kelly, Ph.D. Associate Professor, Department of Economics, University of Wisconsin-River Falls

10 years, 10 months

2
7
0 / 0

Problem with new binary data file format and a question concerning join command

by Pindar

Hi there, I'm using gretl to construct panels and load data in via the great join command - thanks for developing this feature! When I tried to save the final data file in the new gdtb format something went wrong. gretl saves a file but crashes when opening it. Below is the the hansl script to produce the error. In terms of the join command: a compound condition of four and more elements is not supported, or? Have a nice St. Nicholas' Day Leon #**************** set echo off set messages off nulldata 10 --preserve scalar Y1 = 2001 scalar Y2=Y1 + 25 #Level 1 matrix M1 = \ {510\ ;520} string file = "markers_Panel_" sprintf stamp "%d", abs(round(randgen1(z, 0, 1)*randgen1(z, 0, 1)*randgen1(i, 50, 200))) file += stamp file += ".txt" print file outfile "@file" --write loop i=1..rows(M1) --quiet loop j = Y1..Y2 --quiet printf "%g/%g\n", M1[i], j endloop endloop outfile --close scalar N = rows(M1) scalar T = (Y2-Y1+1) scalar counts = N*T nulldata counts --preserve setobs T 1:01 --stacked-time-series series BVNR = M1**ones(T,1) setinfo BVNR --description="Unit ID" series time = ones(N,1)**seq(Y1,Y2)' setinfo time --description="Years" delete index list lGroups = BVNR time discrete lGroups setobs BVNR time --panel-vars sprintf strY1 "%g", Y1 setobs 1 @strY1 --panel-time series time_gretl = $obsdate markers --from-file="@file" store "testPanel.gdt" open "testPanel.gdt" #store "testPanel.gdtb" #open "testPanel.gdtb" #************************

10 years, 10 months

2
2
0 / 0

Nested Logit

by David van Herick

Does anyone know if it is possible to run a nested logit model in gretl? It seems it must be possible somehow with a hansl script or something, but I can't seem to figure it out. I am hoping for an FIML version, but would be happy to even find something that will do a two-step LIML version with multinomial logit models that include logsum parameters.

10 years, 10 months

4
11
0 / 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Gretl-users December 2014