odds ratio and confidence intervals in SPSS

[shawn....@gmail.com]

I am running a logistic regression in SPSS 16.0. I am using smoking
status (smoker vs. non-smoker) as a dependent variable and demographic
variables such as race, sex, age, etc a predictors. I want the odds
ratios and 95% confidence intervals for each level of the predictor
variables. That is, I want the OR's and CI's for each category of
race. Anyone have any ideas?

Thanks,
Shawn

[Bruce Weaver]

The Exp(B) column in the table of coefficients gives you the odds
ratios, and 95% CIs are shown in the next two columns (if you ask
for them). Are you looking for something beyond that?

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/
"When all else fails, RTFM."

[shawn....@gmail.com]

On Sep 28, 8:26 pm, Bruce Weaver <bwea...@lakeheadu.ca> wrote:

> shawn.ache...@gmail.com wrote:
> > I am running a logistic regression in SPSS 16.0. I am using smoking
> > status (smoker vs. non-smoker) as a dependent variable and demographic
> > variables such as race, sex, age, etc a predictors. I want the odds
> > ratios and 95% confidence intervals for each level of the predictor
> > variables. That is, I want the OR's and CI's for each category of
> > race. Anyone have any ideas?
>
> > Thanks,
> > Shawn
>
> The Exp(B) column in the table of coefficients gives you the odds
> ratios, and 95% CIs are shown in the next two columns (if you ask
> for them).  Are you looking for something beyond that?
>
> --
> Bruce Weaver

> bwea...@lakeheadu.cahttp://sites.google.com/a/lakeheadu.ca/bweaver/

> "When all else fails, RTFM."

Thanks for your response. I am looking for the ORs and CIs for the
levels/categories of each predictor variable. I see where it gives me
the OR & CI for race, but I want OR & CI for each category of race
(e.g., arican american, native american, etc), relative to the
comparison group (e.g., caucasian). Any ideas?

[Kylie]

Hi Shawn,

It sounds like you may not have defined 'race' as a categorical
predictor. If using the dialogue box, make sure you go to the
'Categorical' button and select 'race' as a categorical predictor.
'Caucasian' will be your reference category. If using syntax then you
want a subcommand such as:

/CONTRAST (race)=Indicator

Hope this helps.
Cheers,
Kylie.

[shawn....@gmail.com]

Hi Kylie,
I tried that too. Still only gives me the OR & CI for the race
variable as a whole, not the individual categories. I know SAS will
provide this info but alas I don't have access to SAS. Any other
thoughts?

Thanks,
Shawn

[peter m sopp]

shawn....@gmail.com schrieb:

Hello Shawn,
I'm not familiar with dialogue boxes. Therefore: Simply use dummies
(=number of dummies is n-1 because one is the reference group) and
include them in your definition of logistic regression.
Peter

[Bruce Weaver]

You don't need SAS. SPSS produces odds ratios & 95% CIs for k-1 of
the k categories if you give it the right instructions. Here's an
example using one of the sample files that comes with SPSS. Paste the
following example into a syntax window and run it.

* --- Binary Logistic Regression Example --- .

GET FILE='C:\Program Files\SPSS\Samples\1991 U.S. General Social
Survey.sav'.

freq sex.
compute male = (sex EQ 1).
format male (f1.0).
freq male race region.

* Treat REGION and RACE as categorical variables.
* Use last category of REGION and 1st category of RACE
* as reference categories.

LOGISTIC REGRESSION VARIABLES male
/METHOD=ENTER race region
/CONTRAST (region)=Indicator
/CONTRAST (race)=Indicator(1)
/PRINT=CI(95)
/CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).

* Method 2: Create your own indicator varibles .

compute race1 = (race EQ 1).
compute race2 = (race EQ 2).
compute race3 = (race EQ 3).

compute reg1 = (region EQ 1).
compute reg2 = (region EQ 2).
compute reg3 = (region EQ 3).
exe.

format race1 to reg3 (f1.0).

* Include 2 of the 3 indicators for RACE & REGION.
* Omit the indicator for the desired reference category.

LOGISTIC REGRESSION VARIABLES male
/METHOD=ENTER race2 race3 reg1 reg2
/PRINT=CI(95)
/CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5).

* --- End of Example --- .

Notice that there are 2 odds ratios for RACE and 2 for REGION. That
is because each one of them as 3 levels. The omitted 3rd level is the
reference category.

One advantage of the first method above (treating the variables as
categorical rather than creating your own indicator variables) is that
you get a k-1 degree of freedom test for each categorical predictor
variable.

--
Bruce Weaver
bwe...@lakeheadu.ca