On Tue, 27 Dec 2011, Henrique Andrade wrote:
I think I'd found a problem in the Gretl session files. I'm
different kind of models inside a .gretl file (OLS, VAR, and ARIMA)
but she is saying that she can't rebuild some objects.
Using australia.gdt I'm estimating the following models:
ols E const PAU PUS --robust
var 4 PAU PUS E --robust
arima 1 0 1 ; 3
Looking further into the session.xml file (inside "teste.gretl" -
attached in this e-mail), I can see all my models:
<session-model name="Modelo 1" fname="model.1" type="1"
<session-model name="VAR 1" fname="model.2" type="3"
<session-model name="Modelo 2" fname="model.3" type="1"
Is this a bug?
Yes, I'm afraid so. Thanks for the report. The problem was to do
with the VAR, and was caused by an error in the writing of gretl
matrices in XML form. This is now fixed in CVS; I'll put up revised
snapshots when I can, but I'll be in transit today.
An existing session file can be fixed as follows:
(a) Unzip the .gretl file, find the internal file that contains the
VAR, and open it in a text editor.
(b) Look for lines that start with
If such a line ends with, for example,
fix it by adding the missing closing quote. That is,
t2="76> # WRONG
t2="76"> # RIGHT
(c) Rezip the session file, and change its filename extension from
".zip" to ".gretl". It should now open OK in gretl.