Log Reg does not accept /METHOD
Locked
Log Reg does not accept /METHOD
 
|
I am working with Logistic Regression, but the command (v.
18) apparently refuses to accept the /METHOD subcommand. It works fine without
/METHOD. But whenever I include that subcommand, either to run a stepwise run,
or to include multiple ENTER methods, or whatever, the following warning
appears:
The dependent variable is binary. Predictors are interval
measures plus two categorical variables. The general form of the syntax (with
/METHOD included) is as follows: LOGISTIC REGRESSION VARIABLES=yearmom WITH ecolprov urbarur bigtioc
climadem p29 educ single /categorical ecolprov bigtioc /contrast (ecolprov) indicator (1) /contrast (bigtioc) indicator (1) /method =fstep (cond). This works without /METHOD, and fails otherwise. What gives? Thanks in advance for any help. Hector |
Re: Log Reg does not accept /METHOD
|
|
|
Re: Log Reg does not accept /METHOD
|
|
In reply to this post by Hector Maletta
Three may be something else going on here
since you're not getting the error message I would expect, but the METHOD
subcommand should include a list of independent variables. But if that
were the only problem, you'd see an error message that says you need to
specify variables on the METHOD subcommand.
I am working with Logistic Regression, but the command (v. 18) apparently refuses to accept the /METHOD subcommand. It works fine without /METHOD. But whenever I include that subcommand, either to run a stepwise run, or to include multiple ENTER methods, or whatever, the following warning appears:
The dependent variable is binary. Predictors are interval measures plus two categorical variables. The general form of the syntax (with /METHOD included) is as follows: LOGISTIC REGRESSION VARIABLES=yearmom WITH ecolprov urbarur bigtioc climadem p29 educ single /categorical ecolprov bigtioc /contrast (ecolprov) indicator (1) /contrast (bigtioc) indicator (1) /method =fstep (cond). This works without /METHOD, and fails otherwise. What gives? Thanks in advance for any help. Hector |
Re: Log Reg does not accept /METHOD
|
|
In reply to this post by Hector Maletta
The simplest answer is probably to believe
the error message. I think you may have misspelled one of your variable
names, since this works:
dataset close all. input program. loop #i=1 to 100. compute yearmom=rv.bernoulli (.5). do repeat x=ecolprov bigtioc. compute x=trunc(rv.uniform(1,6)). end repeat. do repeat x=urbarur climadem p29 educ single. compute x=rv.normal(50,10). end repeat. end case. end loop. end file. end input program. LOGISTIC REGRESSION VARIABLES=yearmom WITH ecolprov urbarur bigtioc climadem p29 educ single /categorical ecolprov bigtioc /contrast (ecolprov) indicator (1) /contrast (bigtioc) indicator (1) /method =fstep (cond).
Rick, If all variables mentioned in the VARIABLES subcommand are included in the /METHOD subcommand, no variable list is (apparently) needed. However, I have tried the command with and without a variable list included in it, with the same result. And as you point out, the message would be different if that was the origin of the error. I was in fact surprised that my error message does not appear to be an error message but a mere warning (and those ordinarily don’t stop commands to be executed), and also surprised that no variables are named when the warning is that an unknown variable was included. Weird indeed. Dale Glaser might have a point: possibly a bug to be fixed by a patch. But to my knowledge nobody has ever mentioned such a bug. Perhaps some friend at SPSS may clarify this. Hector From: Rick Oliver [mailto:oliverr@...] Sent: Wednesday, August 11, 2010 5:13 PM To: Hector Maletta CC: [hidden email] Subject: Re: Log Reg does not accept /METHOD Three may be something else going on here since you're not getting the error message I would expect, but the METHOD subcommand should include a list of independent variables. But if that were the only problem, you'd see this error message:
I am working with Logistic Regression, but the command (v. 18) apparently refuses to accept the /METHOD subcommand. It works fine without /METHOD. But whenever I include that subcommand, either to run a stepwise run, or to include multiple ENTER methods, or whatever, the following warning appears:
Se certificó que el correo entrante no contiene
virus.
|
Re: Log Reg does not accept /METHOD
|
|
In reply to this post by Hector Maletta
Then I'm stumped. My example, using your
logistic regression syntax, works fine for me in release 18, Do you get
the same error message when you run my example?
Nope. All variable names are right, directly copied from the variable list. Hector De: Rick Oliver [mailto:oliverr@...] Enviado el: Wednesday, August 11, 2010 5:43 PM Para: Hector Maletta CC: [hidden email]; [hidden email] Asunto: RE: Log Reg does not accept /METHOD Ah. The simplest explanation is to believe the error message. I think you probably got one of the variable names wrong. I created some fake data with your variable names, and your syntax runs fine... dataset close all. input program. loop #i=1 to 100. compute yearmom=rv.bernoulli (.5). do repeat x=ecolprov bigtioc. compute x=trunc(rv.uniform(1,6)). end repeat. do repeat x=urbarur climadem p29 educ single. compute x=rv.normal(50,10). end repeat. end case. end loop. end file. end input program. LOGISTIC REGRESSION VARIABLES=yearmom WITH ecolprov urbarur bigtioc climadem p29 educ single /categorical ecolprov bigtioc /contrast (ecolprov) indicator (1) /contrast (bigtioc) indicator (1) /method =fstep (cond).
Rick, If all variables mentioned in the VARIABLES subcommand are included in the /METHOD subcommand, no variable list is (apparently) needed. However, I have tried the command with and without a variable list included in it, with the same result. And as you point out, the message would be different if that was the origin of the error. I was in fact surprised that my error message does not appear to be an error message but a mere warning (and those ordinarily don’t stop commands to be executed), and also surprised that no variables are named when the warning is that an unknown variable was included. Weird indeed. Dale Glaser might have a point: possibly a bug to be fixed by a patch. But to my knowledge nobody has ever mentioned such a bug. Perhaps some friend at SPSS may clarify this. Hector From: Rick Oliver [mailto:oliverr@...] Sent: Wednesday, August 11, 2010 5:13 PM To: Hector Maletta CC: [hidden email] Subject: Re: Log Reg does not accept /METHOD Three may be something else going on here since you're not getting the error message I would expect, but the METHOD subcommand should include a list of independent variables. But if that were the only problem, you'd see this error message:
I am working with Logistic Regression, but the command (v. 18) apparently refuses to accept the /METHOD subcommand. It works fine without /METHOD. But whenever I include that subcommand, either to run a stepwise run, or to include multiple ENTER methods, or whatever, the following warning appears:
Se certificó que el correo entrante no contiene
virus. Se certificó que el correo entrante no contiene
virus.
|
Re: Log Reg does not accept /METHOD - spss case 731024
|
|
In reply to this post by Hector Maletta
Thanks to David and everybody. Problem overcome. It was apparently due to my not
having installed the 18.0.2 patch. I installed the patch and the problem
seemingly disappeared, even if couldn’t find mention of anything related to
this in the 18.0.1 or 18.0.2 patch documentation.
Se certificó
que el correo entrante no contiene virus. Se certificó
que el correo entrante no contiene virus. |
| Free forum by Nabble | Edit this page |