Most of the time, I fail to reconstitute the sessions I have saved. I sometimes manage to open them, but I don't understand what is different from when I can't. Am I missing something about it ?

From: Allin Cottrell <cottrell(a)wfu.edu&gt; Reply-To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu&gt; To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu&gt; Subject: Re: [Gretl-users] sessions reconstitution [2] Date: Mon, 23 Jan 2006 21:06:59 -0500 (EST) On Mon, 23 Jan 2006, Arnaud Bervas wrote: >Most of the time, I fail to reconstitute the sessions I have saved. A second thought: Does anyone on the list happen to know how R's mechanism for saving the binary image of one's workspace works? It may be that a radical redesign of gretl's "session saving" mechanism is in order, along the lines of doing some sort of binary memory dump. Allin Cottrell _______________________________________________ Gretl-users mailing list Gretl-users(a)ricardo.ecn.wfu.edu http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users

Thanx for that Allin. Now I have another problem with exogenous variables. In VAR module (and in VECM too) there is no possibility to specify number of lags of exogenous variables. They are allways included with zero lags. I also tried with creating lags manually and then to include them as exogenous but it's not working. They do not apper in VAR and VECM modules (This is possible in OLS). This means that Gretl threats exogenous as deterministics i.e. as constant and trend i.e. without lags. Is it possible to do that somehow? If not, then I am opting for this option as a new feature! This is realy an usefull thing.

For Jack: I tried to do AIC and BIC tests for lag selection in VAR but the procedure stops here matrix mincrit = { 1.0E20 , 1.0E20 } I think the problem is in numbers in brackets. As I am from Croatia, I tried to change Regional settings on Italian and English but the procedure still remains unfinished. I'll try again, maybe I missed something. Do you have an idea?

It will be great to include it as a feature in next version.

Something that's been floating around in my mind for a while: most people who actively work on the gretl development are Linux users: Allin, Ignacio, Cristian, myself, for instance, are all linux addicts. On the other hand, I have the feeling that most gretl users use windows as their primary environment (this is certainly true of my students). In some cases, this is a nuisance: once a Windows-only user spots a bug, he has no other choice but to report it, hoping it will be fixed soon, even if he had all the necessary skills to fix it himself. That's because up to now no-one (as far as I know) has ever tried to compile gretl under windows. Building gretl under windows is certainly possible. Unfortunately, I have no experience whatsoever with the GNU toolchain for Win32 (nor have I intention of setting up a Windows box for trying, no thanks); but I know for a fact that all the programs you need to build gretl (gcc compiler, make etc.) are available under windows; one implementation (but there may be others) can be found at http://cygwin.org. Has anyone on the list some experience with this?

Sorry, I wasn't quite clear. The purpose is not to have a windows executable; we've got that. The purpose is assembling a recipe for compiling the source under windows.

> Now I have another problem with exogenous variables. In VAR module (and in > VECM too) there is no possibility to specify number of lags of exogenous > variables. They are allways included with zero lags. I also tried with > creating lags manually and then to include them as exogenous but it's not > working. They do not apper in VAR and VECM modules (This is possible in > OLS). This means that Gretl threats exogenous as deterministics i.e. as > constant and trend i.e. without lags.

Short-run workaround: define lagged variables using another name. For example genr xlag1 = x(-1)

In the long run, I still think this could be solved by setting up tools for handling lists in the GUI interface, but that will have to wait a bit.

> I tried to do AIC and BIC tests for lag selection in VAR but the > procedure stops here > matrix mincrit = { 1.0E20 , 1.0E20 }

Sorry, I was assuming you're running the CVS version. If you're using Windows, that should work with the nightly snapshot, which is basically CVS with a 1-day lag. The matrix infrastructure is totally new and the script cannot work with 1.5.0.

From: Allin Cottrell <cottrell(a)wfu.edu&gt; Reply-To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu&gt; To: jack(a)deanovell.unian.it, Gretl list <gretl-users(a)ricardo.ecn.wfu.edu&gt; Subject: VAR dialog (was Re: [Gretl-users] sessions reconstitution [2]) Date: Tue, 24 Jan 2006 11:55:17 -0500 (EST) On Tue, 24 Jan 2006, jack quoted john w: >>Now I have another problem with exogenous variables. In VAR module (and >>in >>VECM too) there is no possibility to specify number of lags of exogenous >>variables. They are allways included with zero lags. I also tried with >>creating lags manually and then to include them as exogenous but it's not >>working. They do not apper in VAR and VECM modules (This is possible in >>OLS). This means that Gretl threats exogenous as deterministics i.e. as >>constant and trend i.e. without lags. The VAR command is one of the older interfaces in the program, and it could do with some updating. In CVS I have added a "--trend" option to the command, with an associated checkbox, "Include a trend", in the VAR dialog box. I have also the renamed the upper right-hand variable selection box to "Exogenous" rather than "Deterministic" variables. However, as you point out, including lags of exogenous variables is problematic. In the original design, the upper box was indeed intended for non-stochastic variables (trend, dummies), where the inclusion of lags would not make sense. And lagged variables are not shown in the left-hand list, for the obvious reason that this could get very confusing, since lags of the endogenous variables are handled automatically. Jack suggests: >Short-run workaround: define lagged variables using another name. For >example > > genr xlag1 = x(-1) Yes, this is a good idea (for the present). You can thereby fool gretl into thinking that "xlag1" is not really a lag, and it will be displayed in the left-hand variable selection list. >In the long run, I still think this could be solved by setting up tools >for handling lists in the GUI interface, but that will have to wait a bit. Agreed, on both counts. Quting john w again: >>I tried to do AIC and BIC tests for lag selection in VAR but the >>procedure stops here >>matrix mincrit = { 1.0E20 , 1.0E20 } Jack: >Sorry, I was assuming you're running the CVS version. If you're using >Windows, that should work with the nightly snapshot, which is basically >CVS with a 1-day lag. The matrix infrastructure is totally new and the >script cannot work with 1.5.0. The Windows snapshot usually follows CVS with a short lag, but there's some discretion here and in fact I haven't made a snapshot lately since there's a lot of new material in CVS that could be unstable. There will be a new snapshot as soon as CVS has settled down (shouldn't be long now). Allin Cottrell _______________________________________________ Gretl-users mailing list Gretl-users(a)ricardo.ecn.wfu.edu http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users