Dear All at EI listserv,
I found the problem. Gauss 5.0 imposes restrictions on data conversions that were not present in earlier versions. In order to convert a character string it is necessary (in 5.0, but not before) to prepend certain defining characters, which the module eicml.src does not do. So the EI modules will not work with Gauss 5.0.
Thanks all,
Paul Sullins
----- Original Message -----
From: Paul Sullins
To: ei(a)latte.harvard.edu
Sent: Thursday, January 23, 2003 2:18 PM
Subject: Problem with EI/CML program
Dear All at EI listserv,
I have a frustrating problem with EI that I hope someone can help/advise me regarding. I am a sociologist of religion at Catholic University, studying the apportionment of religious adherents by region and denomination.
I just bought Gauss LIght 5.0, CML 2.0, and downloaded the latest version of the EI module, but I can't get this set of programs to actually work. Using GAUSS light 5.0, CML 2.0, and EI version 1.8a (possibly; see below), I get the error message
error G0071 : Type mismatch
After it happened with my data, I used the sample data that come with EI, and copied the GAUSS commands to run it directly from the manual. But I get the same message. Here's a copy of my command and output windows:
Commands:
» new;
» library ei;
» clear t x n;
» loadvars sample.asc t x n;
» dbuf=ei(t,x,n,1,1);
Produces output:
----- EI Begins -----
EI Version: 1.63, 2/3/2000
Inputs ok, beginning preliminary estimation...
Likelihood estimation...
d:\program files\gausslt50\src\eicml.src(103) : error G0071 : Type mismatch
(End of output)
Because "type mismatch" indicates the program was expecting a numeric when it got a string, or vice versa, I have tried re-defining the matrices. I've also tried creating a Gauss dataset rather than ascii, bypassing loadvars. No change.
Is anyone else having this problem?
Either A) I'm missing something or B) the EI module doesn't like CML or vice versa.
If it's A, can anyone see my error?
If it's B, does anyone have a copy of the EI archive that you downloaded before 12/10/2002? Perhaps a previous version of the module works.
In fact, if anyone has ANY version of EI that actually works for you, can you send me a copy, to see if it will work with my CML?
Note that, although I downloaded the EI archive on 1/13/03 (and again on 1/17/03 just to be sure), and the web site reports that the latest version, posted 12/10/2002, is 1.8a, the banner message in my Gauss run reports version 1.63. I don't know if this means the new version is actually not on the website, or just that the banner message has not been updated.
Thank you and bless you for any help you can give.
Best wishes,
Paul Sullins
sullins(a)cua.edu
Dear All at EI listserv,
I have a frustrating problem with EI that I hope someone can help/advise me regarding. I am a sociologist of religion at Catholic University, studying the apportionment of religious adherents by region and denomination.
I just bought Gauss LIght 5.0, CML 2.0, and downloaded the latest version of the EI module, but I can't get this set of programs to actually work. Using GAUSS light 5.0, CML 2.0, and EI version 1.8a (possibly; see below), I get the error message
error G0071 : Type mismatch
After it happened with my data, I used the sample data that come with EI, and copied the GAUSS commands to run it directly from the manual. But I get the same message. Here's a copy of my command and output windows:
Commands:
» new;
» library ei;
» clear t x n;
» loadvars sample.asc t x n;
» dbuf=ei(t,x,n,1,1);
Produces output:
----- EI Begins -----
EI Version: 1.63, 2/3/2000
Inputs ok, beginning preliminary estimation...
Likelihood estimation...
d:\program files\gausslt50\src\eicml.src(103) : error G0071 : Type mismatch
(End of output)
Because "type mismatch" indicates the program was expecting a numeric when it got a string, or vice versa, I have tried re-defining the matrices. I've also tried creating a Gauss dataset rather than ascii, bypassing loadvars. No change.
Is anyone else having this problem?
Either A) I'm missing something or B) the EI module doesn't like CML or vice versa.
If it's A, can anyone see my error?
If it's B, does anyone have a copy of the EI archive that you downloaded before 12/10/2002? Perhaps a previous version of the module works.
In fact, if anyone has ANY version of EI that actually works for you, can you send me a copy, to see if it will work with my CML?
Note that, although I downloaded the EI archive on 1/13/03 (and again on 1/17/03 just to be sure), and the web site reports that the latest version, posted 12/10/2002, is 1.8a, the banner message in my Gauss run reports version 1.63. I don't know if this means the new version is actually not on the website, or just that the banner message has not been updated.
Thank you and bless you for any help you can give.
Best wishes,
Paul Sullins
sullins(a)cua.edu