logistic output question

[Maguin, Eugene]

I need to compute confidence intervals for main effects and an interaction combination from a logistic regression. Unlike Regression, which outputs the variance-covariance matrix for the Bs, Logistic outputs the correlation matrix for the Bs. In passing, this seems like a (really) odd design decision but maybe I have something to learn here. First, am I right in thinking that the correlations will not work; that I need to use the covariances? Second, I believe that I will need to use the SEs from the output table to convert the correlations to covariances. Is that true? I haven’t done CIs for a logistic with an interaction and I want to make sure my thinking is correct.

Thanks, Gene Maguin

[Bruce Weaver]

Gene, you could use NOMREG to estimate your model.  Its /PRINT sub-command allows you to print the covariance matrix for the Bs.  

HTH.

Maguin, Eugene wrote

I need to compute confidence intervals for main effects and an interaction combination from a logistic regression. Unlike Regression, which outputs the variance-covariance matrix for the Bs, Logistic outputs the correlation matrix for the Bs. In passing, this seems like a (really) odd design decision but maybe I have something to learn here. First, am I right in thinking that the correlations will not work; that I need to use the covariances? Second, I believe that I will need to use the SEs from the output table to convert the correlations to covariances. Is that true? I haven't done CIs for a logistic with an interaction and I want to make sure my thinking is correct.
Thanks, Gene Maguin

[Ryan]

I'm confused with the question. Doesn't the logistic regression procedure provide confidence intervals for all effects specified in the model? What am I missing?

Ryan

Sent from my iPhone

> On Feb 25, 2014, at 5:35 PM, Bruce Weaver <[hidden email]> wrote:
>
> Gene, you could use NOMREG to estimate your model.  Its /PRINT sub-command
> allows you to print the covariance matrix for the Bs.
>
> HTH.
>
>
>
> Maguin, Eugene wrote
>> I need to compute confidence intervals for main effects and an interaction
>> combination from a logistic regression. Unlike Regression, which outputs
>> the variance-covariance matrix for the Bs, Logistic outputs the
>> correlation matrix for the Bs. In passing, this seems like a (really) odd
>> design decision but maybe I have something to learn here. First, am I
>> right in thinking that the correlations will not work; that I need to use
>> the covariances? Second, I believe that I will need to use the SEs from
>> the output table to convert the correlations to covariances. Is that true?
>> I haven't done CIs for a logistic with an interaction and I want to make
>> sure my thinking is correct.
>> Thanks, Gene Maguin
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question-tp5724626p5724628.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question-tp5724626p5724633.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Yes, it does.  I assumed Gene must be asking about CIs for differences between two fitted values (or some other contrast) that would not appear in the table of coefficients.  But you are right to ask for clarification!  

Bruce

Ryan Black wrote

I'm confused with the question. Doesn't the logistic regression procedure provide confidence intervals for all effects specified in the model? What am I missing?

Ryan

Sent from my iPhone

> On Feb 25, 2014, at 5:35 PM, Bruce Weaver <[hidden email]> wrote:
>
> Gene, you could use NOMREG to estimate your model.  Its /PRINT sub-command
> allows you to print the covariance matrix for the Bs.
>
> HTH.
>
>
>
> Maguin, Eugene wrote
>> I need to compute confidence intervals for main effects and an interaction
>> combination from a logistic regression. Unlike Regression, which outputs
>> the variance-covariance matrix for the Bs, Logistic outputs the
>> correlation matrix for the Bs. In passing, this seems like a (really) odd
>> design decision but maybe I have something to learn here. First, am I
>> right in thinking that the correlations will not work; that I need to use
>> the covariances? Second, I believe that I will need to use the SEs from
>> the output table to convert the correlations to covariances. Is that true?
>> I haven't done CIs for a logistic with an interaction and I want to make
>> sure my thinking is correct.
>> Thanks, Gene Maguin
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question-tp5724626p5724628.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Maguin, Eugene]

Bruce,
Thanks for the note about NomReg. I didn't think to check that routine.

Ryan,
Yes, I see that the logistic output includes a CI option (and the ability to specify the significance level seems to be a very nice feature).  My starting point was a model that included an interaction term and in my results text I planned to state the OR and its CI for each combination of IV values. Thus the need to compute a composite SE. I expected to find the cov matrix.

Gene

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Ryan Black
Sent: Tuesday, February 25, 2014 7:15 PM
To: [hidden email]
Subject: Re: logistic output question

I'm confused with the question. Doesn't the logistic regression procedure provide confidence intervals for all effects specified in the model? What am I missing?

Ryan

Sent from my iPhone

> On Feb 25, 2014, at 5:35 PM, Bruce Weaver <[hidden email]> wrote:
>
> Gene, you could use NOMREG to estimate your model.  Its /PRINT
> sub-command allows you to print the covariance matrix for the Bs.
>
> HTH.
>
>
>
> Maguin, Eugene wrote
>> I need to compute confidence intervals for main effects and an
>> interaction combination from a logistic regression. Unlike
>> Regression, which outputs the variance-covariance matrix for the Bs,
>> Logistic outputs the correlation matrix for the Bs. In passing, this
>> seems like a (really) odd design decision but maybe I have something
>> to learn here. First, am I right in thinking that the correlations
>> will not work; that I need to use the covariances? Second, I believe
>> that I will need to use the SEs from the output table to convert the correlations to covariances. Is that true?
>> I haven't done CIs for a logistic with an interaction and I want to
>> make sure my thinking is correct.
>> Thanks, Gene Maguin
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question
> -tp5724626p5724628.html Sent from the SPSSX Discussion mailing list
> archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except
> the command. To leave the list, send the command SIGNOFF SPSSX-L For a
> list of commands to manage subscriptions, send the command INFO
> REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Hi Gene.  NOMREG also has a /TEST sub-command which you can use to generate fitted values at selected combinations of your predictors, and differences between two fitted values, etc.  Its output should include standard errors and confidence intervals.  

Alternatively, you could try one of the macros I wrote *before* I realized that I could have used NOMREG and /TEST for logistic regression models, or UNIANOVA + /LMATRIX for OLS models.  ;-)  See the material at the link below:

  https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/spss/olscomp_mlecomp

HTH.

Bruce

Maguin, Eugene wrote

Bruce,
Thanks for the note about NomReg. I didn't think to check that routine.

Ryan,
Yes, I see that the logistic output includes a CI option (and the ability to specify the significance level seems to be a very nice feature).  My starting point was a model that included an interaction term and in my results text I planned to state the OR and its CI for each combination of IV values. Thus the need to compute a composite SE. I expected to find the cov matrix.

Gene

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Ryan Black
Sent: Tuesday, February 25, 2014 7:15 PM
To: [hidden email]
Subject: Re: logistic output question

I'm confused with the question. Doesn't the logistic regression procedure provide confidence intervals for all effects specified in the model? What am I missing?

Ryan

Sent from my iPhone

> On Feb 25, 2014, at 5:35 PM, Bruce Weaver <[hidden email]> wrote:
>
> Gene, you could use NOMREG to estimate your model.  Its /PRINT
> sub-command allows you to print the covariance matrix for the Bs.
>
> HTH.
>
>
>
> Maguin, Eugene wrote
>> I need to compute confidence intervals for main effects and an
>> interaction combination from a logistic regression. Unlike
>> Regression, which outputs the variance-covariance matrix for the Bs,
>> Logistic outputs the correlation matrix for the Bs. In passing, this
>> seems like a (really) odd design decision but maybe I have something
>> to learn here. First, am I right in thinking that the correlations
>> will not work; that I need to use the covariances? Second, I believe
>> that I will need to use the SEs from the output table to convert the correlations to covariances. Is that true?
>> I haven't done CIs for a logistic with an interaction and I want to
>> make sure my thinking is correct.
>> Thanks, Gene Maguin
>
>
>
>
>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question
> -tp5724626p5724628.html Sent from the SPSSX Discussion mailing list
> archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except
> the command. To leave the list, send the command SIGNOFF SPSSX-L For a
> list of commands to manage subscriptions, send the command INFO
> REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Maguin, Eugene]

Bruce, I noted you comment about the SE covs being obtainable from NomReg (NR) and out curiosity I ran the same model in both logistic regression (LR) and requested SE corrs and covs. Since you wrote that macro, I am curious if you noticed or have noticed this? However in what follows I'm posing the observation to the listserv and NOT to you specifically.

So, same model run in LR and NR. Corresponding parameter estimates are within 1% of each other. But SE corrs! Apparently startling, but I might have done something wrong or just simply not understand something. Anway.

LR.
        Const   M(1)    A(1)
Const   1.000   .242    -.162
M(1)    .242    1.000   -.097
A(1)    -.162   -.097   1.000

NR.
               Intercept        [M=0]   [M=1]   [A=0]   [A=1]
Intercept       1       -.408    .b     -.648   .b
[M=.00] -.408   1       .b      -.097   .b
[M=1.00]        .b      .b      .b      .b      .b
[A=.00 ]        -.648   -.097   .b      1       .b
[A=1.00]        .b      .b      .b      .b      .b
Note. bRedundant parameter estimate.

The syntax is
LOGISTIC REGRESSION R WITH M A/CATEGORICAL M A/CONTRAST(M)=SIMPLE(1)/
   CONTRAST(A)=SIMPLE(1)/ENTER M A/ PRINT=DEFAULT CORR.

NOMREG R BY M A/MODEL=M A/PRINT PARAMETER CORB COVB.

Gene Maguikn

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver
Sent: Wednesday, February 26, 2014 1:48 PM
To: [hidden email]
Subject: Re: logistic output question

Hi Gene.  NOMREG also has a /TEST sub-command which you can use to generate fitted values at selected combinations of your predictors, and differences between two fitted values, etc.  Its output should include standard errors and confidence intervals.

/Alternatively/, you could try one of the macros I wrote *before* I realized that I could have used NOMREG and /TEST for logistic regression models, or UNIANOVA + /LMATRIX for OLS models.  ;-)  See the material at the link
below:


https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/spss/olscomp_mlecomp

HTH.

Bruce



Maguin, Eugene wrote

> Bruce,
> Thanks for the note about NomReg. I didn't think to check that routine.
>
> Ryan,
> Yes, I see that the logistic output includes a CI option (and the
> ability to specify the significance level seems to be a very nice
> feature).  My starting point was a model that included an interaction
> term and in my results text I planned to state the OR and its CI for
> each combination of IV values. Thus the need to compute a composite
> SE. I expected to find the cov matrix.
>
> Gene
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:

> SPSSX-L@.UGA

> ] On Behalf Of Ryan Black
> Sent: Tuesday, February 25, 2014 7:15 PM
> To:

> SPSSX-L@.UGA

> Subject: Re: logistic output question
>
> I'm confused with the question. Doesn't the logistic regression
> procedure provide confidence intervals for all effects specified in
> the model? What am I missing?
>
> Ryan
>
> Sent from my iPhone
>
>> On Feb 25, 2014, at 5:35 PM, Bruce Weaver &lt;

> bruce.weaver@

> &gt; wrote:
>>
>> Gene, you could use NOMREG to estimate your model.  Its /PRINT
>> sub-command allows you to print the covariance matrix for the Bs.
>>
>> HTH.
>>
>>
>>
>> Maguin, Eugene wrote
>>> I need to compute confidence intervals for main effects and an
>>> interaction combination from a logistic regression. Unlike
>>> Regression, which outputs the variance-covariance matrix for the Bs,
>>> Logistic outputs the correlation matrix for the Bs. In passing, this
>>> seems like a (really) odd design decision but maybe I have something
>>> to learn here. First, am I right in thinking that the correlations
>>> will not work; that I need to use the covariances? Second, I believe
>>> that I will need to use the SEs from the output table to convert the
>>> correlations to covariances. Is that true?
>>> I haven't done CIs for a logistic with an interaction and I want to
>>> make sure my thinking is correct.
>>> Thanks, Gene Maguin
>>
>>
>>
>>
>>
>> -----
>> --
>> Bruce Weaver
>>

> bweaver@

>> http://sites.google.com/a/lakeheadu.ca/bweaver/
>>
>> "When all else fails, RTFM."
>>
>> NOTE: My Hotmail account is not monitored regularly.
>> To send me an e-mail, please use the address shown above.
>>
>> --
>> View this message in context:
>> http://spssx-discussion.1045642.n5.nabble.com/logistic-output-questio
>> n -tp5724626p5724628.html Sent from the SPSSX Discussion mailing list
>> archive at Nabble.com.
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>>

> LISTSERV@.UGA

>  (not to SPSSX-L), with no body text except
>> the command. To leave the list, send the command SIGNOFF SPSSX-L For
>> a list of commands to manage subscriptions, send the command INFO
>> REFCARD
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (not to SPSSX-L), with no body text except the command. To leave the
> list, send the command SIGNOFF SPSSX-L For a list of commands to
> manage subscriptions, send the command INFO REFCARD
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (not to SPSSX-L), with no body text except the command. To leave the
> list, send the command SIGNOFF SPSSX-L For a list of commands to
> manage subscriptions, send the command INFO REFCARD





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question-tp5724626p5724648.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

Hi Gene.  I'm at home now, and don't have SPSS on this machine, so cannot test.  But here are LR and NOMREG commands that I used to run the same model when testing NOMREG a couple years ago.  They use the bankloan.sav data file that comes with SPSS (look in your "sample data" folder, or whatever it is called).


LOGISTIC REGRESSION VARIABLES default
  /METHOD=ENTER educ employ educ*employ
  /CONTRAST (educ)=Indicator(1)
  /PRINT=CI(95)
  /SAVE=PRED(predprob) .

NOMREG
  default  (BASE=FIRST ORDER=ASCENDING) BY educ WITH employ
  /INTERCEPT =INCLUDE
  /MODEL = educ employ educ*employ
  /TEST "FV1: EDUC=1, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  1 0 0 0   10  10 0 0 0;
     ALL  1  1 0 0 0    0   0 0 0 0;
     ALL  0  0 0 0 0   10  10 0 0 0;
     ALL  0  0 0 0 0  -10 -10 0 0 0
  /TEST "FV1: EDUC=2, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0  1 0 0   10  0  10 0 0;
     ALL  1  1  0 0 0    0  0   0 0 0;
     ALL  0 -1  1 0 0   10  0  10 0 0;
     ALL  0  1 -1 0 0  -10  0 -10 0 0
  /TEST "FV1: EDUC=3, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0 0  1 0   10  0 0  10 0;
     ALL  1  1 0  0 0    0  0 0   0 0;
     ALL  0 -1 0  1 0   10  0 0  10 0;
     ALL  0  1 0 -1 0  -10  0 0 -10 0  
  /TEST "FV1: EDUC=4, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0 0 0  1   10 0 0 0  10;
     ALL  1  1 0 0  0    0 0 0 0   0;
     ALL  0 -1 0 0  1   10 0 0 0  10;  
     ALL  0  1 0 0 -1  -10 0 0 0 -10  
  /PRINT = PARAMETER SUMMARY LRT CPS STEP MFI
.

Notice the "(BASE=FIRST ORDER=ASCENDING)".  If memory serves, I needed to do that because the default for NOMREG was to treat the highest value of the DV as the reference category.  Give that a try, and see if it makes thing match up better.  

HTH.
Bruce

[Maguin, Eugene]

Bruce, thank you for your reply. I found that it helps to run the same model through all routines. NomReg (and GenLin) seem to use indicator coding for contrasts and when I used that instead of simple coding for Logistic Regression, I get the same numbers. What a surprise! I do want to work my way through your example below and your macro routine so I understand what's beneath the surface. Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver
Sent: Thursday, February 27, 2014 7:01 PM
To: [hidden email]
Subject: Re: logistic output question

Hi Gene.  I'm at home now, and don't have SPSS on this machine, so cannot test.  But here are LR and NOMREG commands that I used to run the same model when testing NOMREG a couple years ago.  They use the bankloan.sav data file that comes with SPSS (look in your "sample data" folder, or whatever it is called).


LOGISTIC REGRESSION VARIABLES default
  /METHOD=ENTER educ employ educ*employ
  /CONTRAST (educ)=Indicator(1)
  /PRINT=CI(95)
  /SAVE=PRED(predprob) .

NOMREG
  default  (BASE=FIRST ORDER=ASCENDING) BY educ WITH employ
  /INTERCEPT =INCLUDE
  /MODEL = educ employ educ*employ
  /TEST "FV1: EDUC=1, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  1 0 0 0   10  10 0 0 0;
     ALL  1  1 0 0 0    0   0 0 0 0;
     ALL  0  0 0 0 0   10  10 0 0 0;
     ALL  0  0 0 0 0  -10 -10 0 0 0
  /TEST "FV1: EDUC=2, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0  1 0 0   10  0  10 0 0;
     ALL  1  1  0 0 0    0  0   0 0 0;
     ALL  0 -1  1 0 0   10  0  10 0 0;
     ALL  0  1 -1 0 0  -10  0 -10 0 0
  /TEST "FV1: EDUC=3, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0 0  1 0   10  0 0  10 0;
     ALL  1  1 0  0 0    0  0 0   0 0;
     ALL  0 -1 0  1 0   10  0 0  10 0;
     ALL  0  1 0 -1 0  -10  0 0 -10 0
  /TEST "FV1: EDUC=4, EMPLOY=10; FV2: EDUC=1, EMPLOY=0"
     ALL  1  0 0 0  1   10 0 0 0  10;
     ALL  1  1 0 0  0    0 0 0 0   0;
     ALL  0 -1 0 0  1   10 0 0 0  10;
     ALL  0  1 0 0 -1  -10 0 0 0 -10
  /PRINT = PARAMETER SUMMARY LRT CPS STEP MFI .

Notice the "(BASE=FIRST ORDER=ASCENDING)".  If memory serves, I needed to do that because the default for NOMREG was to treat the highest value of the DV as the reference category.  Give that a try, and see if it makes thing match up better.

HTH.
Bruce




-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-output-question-tp5724626p5724665.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD