Difference between Simple and Indicator in logistic regression

[devoidx]

Hi Guys, yet another thing that I am not too sure about in spss, I am trying to run a binary logistic regression and in defining my categorial independent variable , I am unsure how differently the results of indicator and Simple contrasts are interpreted.... I know that Simple compares each group with the reference category so the Odds ratio that I get for each group of my categorical variable is in comparison with the reference category,,,,but what are the Odds ratios indicates if i choose indicator as the method of contrast?  



Thanks as always.

[devoidx]

I've been reading about this  and can't really find how the output will be different if i select simple vs. indicator...exhausted my reading options..if anyone can shed light on this for me, will be much appreciated.

[Maguin, Eugene]

Have you looked standard reference texts yet? It doesn't matter whether the text is about Gauss (normal) regression or logistic regression because your question is about the IVs. And, here is a website link that may be helpful

http://www.ats.ucla.edu/stat/sas/webbooks/reg/chapter5/sasreg5.htm

Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of devoidx
Sent: Wednesday, October 09, 2013 3:39 PM
To: [hidden email]
Subject: Re: Difference between Simple and Indicator in logistic regression

I've been reading about this  and can't really find how the output will be different if i select simple vs. indicator...exhausted my reading options..if anyone can shed light on this for me, will be much appreciated.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Difference-between-Simple-and-Indicator-in-logistic-regression-tp5722455p5722467.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]

I just tried this quick example:

NEW FILE.
DATASET CLOSE all.
DATA LIST list / marital Male Kount (3f5.0).
BEGIN DATA
1 0 727
1 1 619
2 0 232
2 1 51
3 0 258
3 1 188
4 0 53
4 1 40
5 0 329
5 1 334
END DATA.

WEIGHT by Kount.

LOGISTIC REGRESSION VARIABLES male
  /METHOD=ENTER marital
  /CONTRAST (marital)=Indicator
  /PRINT=CI(95)
  /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).

LOGISTIC REGRESSION VARIABLES male
  /METHOD=ENTER marital
  /CONTRAST (marital)=Simple
  /PRINT=CI(95)
  /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).


Here is the Categorical Variable Coding table for the first model with Indicator contrast type:

Marital    P1 to P4
1            1 0 0 0
2            0 1 0 0
3            0 0 1 0
4            0 0 0 1
5            0 0 0 0

And now for the second model with Simple contrast type:

Marital    P1 to P4
1             .8  -.2  -.2  -.2
2            -.2   .8  -.2  -.2  
3            -.2  -.2   .8  -.2
4            -.2  -.2  -.2   .8
5            -.2  -.2  -.2  -.2

Finally, from the FM entry for Logistic Regression > CONTRAST Subcommand:

INDICATOR(refcat).  Indicator variables. Contrasts indicate the presence or absence of category membership. By default, refcat is the last category (represented in the contrast matrix as a row of zeros). To omit a category (other than the last category), specify the sequence number of the omitted category (which is not necessarily the same as its value) in parentheses after the keyword INDICATOR.

SIMPLE(refcat).  Each category of the independent variable (except the last category) is compared to the last category. To use a category other than the last as the omitted reference category, specify its sequence number (which is not necessarily the same as its value) in parentheses following the
keyword SIMPLE.

I understand the OP's confusion:  The description for Simple does not seem to match the coefficients in the Categorical Variables Coding table!

devoidx wrote

I've been reading about this  and can't really find how the output will be different if i select simple vs. indicator...exhausted my reading options..if anyone can shed light on this for me, will be much appreciated.

[Bruce Weaver]

See also:  

http://listserv.uga.edu/cgi-bin/wa?A2=ind9912&L=spssx-l&P=12247
http://www.ats.ucla.edu/stat/spss/library/contrast.htm

On the first of those pages, notice this advice from John Hendrickx:  

    "The bottom line: use the indicator contrast. Same estimates, easier predicted values."

I concur!

Bruce Weaver wrote

I just tried this quick example:

NEW FILE.
DATASET CLOSE all.
DATA LIST list / marital Male Kount (3f5.0).
BEGIN DATA
1 0 727
1 1 619
2 0 232
2 1 51
3 0 258
3 1 188
4 0 53
4 1 40
5 0 329
5 1 334
END DATA.

WEIGHT by Kount.

LOGISTIC REGRESSION VARIABLES male
  /METHOD=ENTER marital
  /CONTRAST (marital)=Indicator
  /PRINT=CI(95)
  /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).

LOGISTIC REGRESSION VARIABLES male
  /METHOD=ENTER marital
  /CONTRAST (marital)=Simple
  /PRINT=CI(95)
  /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).


Here is the Categorical Variable Coding table for the first model with Indicator contrast type:

Marital    P1 to P4
1            1 0 0 0
2            0 1 0 0
3            0 0 1 0
4            0 0 0 1
5            0 0 0 0

And now for the second model with Simple contrast type:

Marital    P1 to P4
1             .8  -.2  -.2  -.2
2            -.2   .8  -.2  -.2  
3            -.2  -.2   .8  -.2
4            -.2  -.2  -.2   .8
5            -.2  -.2  -.2  -.2

Finally, from the FM entry for Logistic Regression > CONTRAST Subcommand:

INDICATOR(refcat).  Indicator variables. Contrasts indicate the presence or absence of category membership. By default, refcat is the last category (represented in the contrast matrix as a row of zeros). To omit a category (other than the last category), specify the sequence number of the omitted category (which is not necessarily the same as its value) in parentheses after the keyword INDICATOR.

SIMPLE(refcat).  Each category of the independent variable (except the last category) is compared to the last category. To use a category other than the last as the omitted reference category, specify its sequence number (which is not necessarily the same as its value) in parentheses following the
keyword SIMPLE.

I understand the OP's confusion:  The description for Simple does not seem to match the coefficients in the Categorical Variables Coding table!

devoidx wrote

I've been reading about this  and can't really find how the output will be different if i select simple vs. indicator...exhausted my reading options..if anyone can shed light on this for me, will be much appreciated.

[devoidx]

Thanks a bunch Bruce, yeah I am gonna stick with indicator since it is the default anyways but was just curious, I'll post back if i come across some literature that describes the difference between these two contrasts.

By the way Bruce, did you get the same EXP(B) [odds ratio] values in both contrast models? that is the part that i was mostly curious about, and whether the odds ratios are interpreted differently in indicator vs. simple. I would try your sample code myself but my spss is tied up right now and probably will be till tomorrow morning around noon.

Thanks again!

[David Marso]

"I would try your sample code myself but my spss is tied up right now and probably will be till tomorrow morning around noon."

NOTE:  One can launch multiple sessions of SPSS!

devoidx wrote

Thanks a bunch Bruce, yeah I am gonna stick with indicator since it is the default anyways but was just curious, I'll post back if i come across some literature that describes the difference between these two contrasts.

By the way Bruce, did you get the same EXP(B) [odds ratio] values in both contrast models? that is the part that i was mostly curious about, and whether the odds ratios are interpreted differently in indicator vs. simple. I would try your sample code myself but my spss is tied up right now and probably will be till tomorrow morning around noon.

Thanks again!

[Bruce Weaver]

From the first model (with Indicator contrasts):

   B             SE
-0.176 0.095  Marital(1)
-1.530 0.173  Marital(2)
-0.332 0.123  Marital(3)
-0.296 0.223  Marital(4)
0.015 0.078  Constant

From the second model (with Simple contrasts):

   B             SE
-0.176 0.095  Marital(1)
-1.530 0.173  Marital(2)
-0.332 0.123  Marital(3)
-0.296 0.223  Marital(4)
-0.452 0.059  Constant

So the only difference in the parameter estimates is for the Constant.

devoidx wrote

Thanks a bunch Bruce, yeah I am gonna stick with indicator since it is the default anyways but was just curious, I'll post back if i come across some literature that describes the difference between these two contrasts.

By the way Bruce, did you get the same EXP(B) [odds ratio] values in both contrast models? that is the part that i was mostly curious about, and whether the odds ratios are interpreted differently in indicator vs. simple. I would try your sample code myself but my spss is tied up right now and probably will be till tomorrow morning around noon.

Thanks again!

[devoidx]

Gotchya, Thanks Bruce.