f



comparison between groups on nominal variables

How does one compare groups on nominal or ordinal variables? I want to know if the groups show significant differences in gender and education. How can one do it on SPSS?

0
snowdenisinrussia
8/3/2016 9:40:51 PM
comp.soft-sys.stat.spss 5679 articles. 0 followers. Post Follow

9 Replies
326 Views

Similar Articles

[PageSpeed] 15

On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.com
wrote:

>How does one compare groups on nominal or ordinal variables? I want to know if the groups show significant differences in gender and education. How can one do it on SPSS?

Google for the UCLA tutorial on using SPSS in data analysis. 

The names of SPSS procedures often match the names of the
tests, such as t-test or ANOVA.  

-- 
Rich Ulrich 
0
Rich
8/4/2016 6:18:41 AM
quinta-feira, 4 de Agosto de 2016 =C3=A0s 07:18:47 UTC+1, Rich Ulrich escre=
veu:
> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.com
> wrote:
>=20
> >How does one compare groups on nominal or ordinal variables? I want to k=
now if the groups show significant differences in gender and education. How=
 can one do it on SPSS?
>=20
> Google for the UCLA tutorial on using SPSS in data analysis.=20
>=20
> The names of SPSS procedures often match the names of the
> tests, such as t-test or ANOVA. =20
>=20
> --=20
> Rich Ulrich



I tried to conduct a Goodness of fit Chi-Squared but instead of comparing t=
he gender proportions between groups it treated the gender variable indepen=
dently of the groups which is not what I want to do. Am I doing something w=
rong?
0
snowdenisinrussia
8/4/2016 8:28:17 AM
On Thu, 4 Aug 2016 01:28:17 -0700 (PDT), snowdenisinrussia@gmail.com
wrote:

>quinta-feira, 4 de Agosto de 2016 às 07:18:47 UTC+1, Rich Ulrich escreveu:
>> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.com
>> wrote:
>> 
>> >How does one compare groups on nominal or ordinal variables? I want to know if the groups show significant differences in gender and education. How can one do it on SPSS?
>> 
>> Google for the UCLA tutorial on using SPSS in data analysis. 
>> 
>> The names of SPSS procedures often match the names of the
>> tests, such as t-test or ANOVA.  
>> 
>> -- 
>> Rich Ulrich
>
>
>
>I tried to conduct a Goodness of fit Chi-Squared but instead of comparing the gender proportions between groups it treated the gender variable independently of the groups which is not what I want to do. Am I doing something wrong?

Okay.  The procedures often match the names of tests.  On the 
other hand, there are a lot of occasions where you might 
apply an F-statistic or a chi-squared, or "goodness-of-fit", 
and you need an orientation to know which test meets your
circumstances.  

If you can't take a course, at least read the tutorial. 

It sounds to me like you wanted a "contingency table" for two
variables but you called upon the "goodness-of-fit" that looked at
each variable alone to see if its cells were all equal.  You can get
surer advice if you show the actual syntax that you ran.  (A good
practice for everyone is to "paste syntax" from the click-and-choose
interface; save your runs for study (at first) and documentation.

-- 
Rich Ulrich 

0
Rich
8/4/2016 4:37:09 PM
quinta-feira, 4 de Agosto de 2016 =C3=A0s 17:37:15 UTC+1, Rich Ulrich escre=
veu:
> On Thu, 4 Aug 2016 01:28:17 -0700 (PDT), snowdenisinrussia@gmail.com
> wrote:
>=20
> >quinta-feira, 4 de Agosto de 2016 =C3=A0s 07:18:47 UTC+1, Rich Ulrich es=
creveu:
> >> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.com
> >> wrote:
> >>=20
> >> >How does one compare groups on nominal or ordinal variables? I want t=
o know if the groups show significant differences in gender and education. =
How can one do it on SPSS?
> >>=20
> >> Google for the UCLA tutorial on using SPSS in data analysis.=20
> >>=20
> >> The names of SPSS procedures often match the names of the
> >> tests, such as t-test or ANOVA. =20
> >>=20
> >> --=20
> >> Rich Ulrich
> >
> >
> >
> >I tried to conduct a Goodness of fit Chi-Squared but instead of comparin=
g the gender proportions between groups it treated the gender variable inde=
pendently of the groups which is not what I want to do. Am I doing somethin=
g wrong?
>=20
> Okay.  The procedures often match the names of tests.  On the=20
> other hand, there are a lot of occasions where you might=20
> apply an F-statistic or a chi-squared, or "goodness-of-fit",=20
> and you need an orientation to know which test meets your
> circumstances. =20
>=20
> If you can't take a course, at least read the tutorial.=20
>=20
> It sounds to me like you wanted a "contingency table" for two
> variables but you called upon the "goodness-of-fit" that looked at
> each variable alone to see if its cells were all equal.  You can get
> surer advice if you show the actual syntax that you ran.  (A good
> practice for everyone is to "paste syntax" from the click-and-choose
> interface; save your runs for study (at first) and documentation.
>=20
> --=20
> Rich Ulrich


I have been reading some of the articles from the UCLA website and I have t=
o thank you for your suggestion.
I do not know how to provide the syntax. So I can just try to give you more=
 information if that is helpful. Basically I want to see if the three group=
s I have differ in gender since I want them to be similar is some variables=
 such as gender, SES, etc. Thus, all I want to verify is if there are diffe=
rences between these groups for those variables. I treated SES as an ordina=
l variable as there is an intrinsic order between the different levels. How=
ever, I am struggling to find the best way of conducting the differences be=
tween gender. I think the most suitable test would be the chi-squared test =
of independence, but I am not 100% certain.

0
snowdenisinrussia
8/4/2016 10:33:16 PM
On Thu, 4 Aug 2016 15:33:16 -0700 (PDT), snowdenisinrussia@gmail.com
wrote:

>quinta-feira, 4 de Agosto de 2016 às 17:37:15 UTC+1, Rich Ulrich escreveu:
>> On Thu, 4 Aug 2016 01:28:17 -0700 (PDT), snowdenisinrussia@gmail.com
>> wrote:
>> 
>> >quinta-feira, 4 de Agosto de 2016 às 07:18:47 UTC+1, Rich Ulrich escreveu:
>> >> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.com
>> >> wrote:
>> >> 
>> >> >How does one compare groups on nominal or ordinal variables? I want to know if the groups show significant differences in gender and education. How can one do it on SPSS?
>> >> 
>> >> Google for the UCLA tutorial on using SPSS in data analysis. 
>> >> 
>> >> The names of SPSS procedures often match the names of the
>> >> tests, such as t-test or ANOVA.  
>> >> 
>> >> -- 
>> >> Rich Ulrich
>> >
>> >
>> >
>> >I tried to conduct a Goodness of fit Chi-Squared but instead of comparing the gender proportions between groups it treated the gender variable independently of the groups which is not what I want to do. Am I doing something wrong?
>> 
>> Okay.  The procedures often match the names of tests.  On the 
>> other hand, there are a lot of occasions where you might 
>> apply an F-statistic or a chi-squared, or "goodness-of-fit", 
>> and you need an orientation to know which test meets your
>> circumstances.  
>> 
>> If you can't take a course, at least read the tutorial. 
>> 
>> It sounds to me like you wanted a "contingency table" for two
>> variables but you called upon the "goodness-of-fit" that looked at
>> each variable alone to see if its cells were all equal.  You can get
>> surer advice if you show the actual syntax that you ran.  (A good
>> practice for everyone is to "paste syntax" from the click-and-choose
>> interface; save your runs for study (at first) and documentation.
>> 
>> -- 
>> Rich Ulrich
>
>
>I have been reading some of the articles from the UCLA website and I have to thank you for your suggestion.
>I do not know how to provide the syntax. So I can just try to give you more information if that is helpful. Basically I want to see if the three groups I have differ in gender since I want them to be similar is some variables such as gender, SES, etc. Thus, all I want to verify is if there are differences between these groups for
those variables. I treated SES as an ordinal variable as there is an intrinsic order between the different levels. However, I am struggling to find the best way of conducting the differences between gender. I think the most suitable test would be the chi-squared test of independence, but I am not 100% certain.

If you are happy with reporting the average score for
SES, then SES can be treated as a continuous variable 
in an ANOVA.  That is fairly conventional. 

The usual test for a dichotomy like sex would be the
contingency table.  The test will be almost the same
p-value as looking at "average sex" in an ANOVA. 

-- 
Rich Ulrich 

0
Rich
8/5/2016 6:47:25 AM
sexta-feira, 5 de Agosto de 2016 =C3=A0s 07:47:31 UTC+1, Rich Ulrich escrev=
eu:
> On Thu, 4 Aug 2016 15:33:16 -0700 (PDT), snowdenisinrussia@gmail.com
> wrote:
>=20
> >quinta-feira, 4 de Agosto de 2016 =C3=A0s 17:37:15 UTC+1, Rich Ulrich es=
creveu:
> >> On Thu, 4 Aug 2016 01:28:17 -0700 (PDT), snowdenisinrussia@gmail.com
> >> wrote:
> >>=20
> >> >quinta-feira, 4 de Agosto de 2016 =C3=A0s 07:18:47 UTC+1, Rich Ulrich=
 escreveu:
> >> >> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.co=
m
> >> >> wrote:
> >> >>=20
> >> >> >How does one compare groups on nominal or ordinal variables? I wan=
t to know if the groups show significant differences in gender and educatio=
n. How can one do it on SPSS?
> >> >>=20
> >> >> Google for the UCLA tutorial on using SPSS in data analysis.=20
> >> >>=20
> >> >> The names of SPSS procedures often match the names of the
> >> >> tests, such as t-test or ANOVA. =20
> >> >>=20
> >> >> --=20
> >> >> Rich Ulrich
> >> >
> >> >
> >> >
> >> >I tried to conduct a Goodness of fit Chi-Squared but instead of compa=
ring the gender proportions between groups it treated the gender variable i=
ndependently of the groups which is not what I want to do. Am I doing somet=
hing wrong?
> >>=20
> >> Okay.  The procedures often match the names of tests.  On the=20
> >> other hand, there are a lot of occasions where you might=20
> >> apply an F-statistic or a chi-squared, or "goodness-of-fit",=20
> >> and you need an orientation to know which test meets your
> >> circumstances. =20
> >>=20
> >> If you can't take a course, at least read the tutorial.=20
> >>=20
> >> It sounds to me like you wanted a "contingency table" for two
> >> variables but you called upon the "goodness-of-fit" that looked at
> >> each variable alone to see if its cells were all equal.  You can get
> >> surer advice if you show the actual syntax that you ran.  (A good
> >> practice for everyone is to "paste syntax" from the click-and-choose
> >> interface; save your runs for study (at first) and documentation.
> >>=20
> >> --=20
> >> Rich Ulrich
> >
> >
> >I have been reading some of the articles from the UCLA website and I hav=
e to thank you for your suggestion.
> >I do not know how to provide the syntax. So I can just try to give you m=
ore information if that is helpful. Basically I want to see if the three gr=
oups I have differ in gender since I want them to be similar is some variab=
les such as gender, SES, etc. Thus, all I want to verify is if there are di=
fferences between these groups for
> those variables. I treated SES as an ordinal variable as there is an intr=
insic order between the different levels. However, I am struggling to find =
the best way of conducting the differences between gender. I think the most=
 suitable test would be the chi-squared test of independence, but I am not =
100% certain.
>=20
> If you are happy with reporting the average score for
> SES, then SES can be treated as a continuous variable=20
> in an ANOVA.  That is fairly conventional.=20
>=20
> The usual test for a dichotomy like sex would be the
> contingency table.  The test will be almost the same
> p-value as looking at "average sex" in an ANOVA.=20
>=20
> --=20
> Rich Ulrich

I did "crosstabs" and chose pearson's chi-square and Phi and Cramer's V on =
statistics. However, I am still struggling when seeing videos of people usi=
ng this test for relationship between variables when I want to use it for d=
ifferences between groups.
0
catiamargarid
8/6/2016 5:07:49 AM
sexta-feira, 5 de Agosto de 2016 =C3=A0s 07:47:31 UTC+1, Rich Ulrich escrev=
eu:
> On Thu, 4 Aug 2016 15:33:16 -0700 (PDT), snowdenisinrussia@gmail.com
> wrote:
>=20
> >quinta-feira, 4 de Agosto de 2016 =C3=A0s 17:37:15 UTC+1, Rich Ulrich es=
creveu:
> >> On Thu, 4 Aug 2016 01:28:17 -0700 (PDT), snowdenisinrussia@gmail.com
> >> wrote:
> >>=20
> >> >quinta-feira, 4 de Agosto de 2016 =C3=A0s 07:18:47 UTC+1, Rich Ulrich=
 escreveu:
> >> >> On Wed, 3 Aug 2016 14:40:51 -0700 (PDT), snowdenisinrussia@gmail.co=
m
> >> >> wrote:
> >> >>=20
> >> >> >How does one compare groups on nominal or ordinal variables? I wan=
t to know if the groups show significant differences in gender and educatio=
n. How can one do it on SPSS?
> >> >>=20
> >> >> Google for the UCLA tutorial on using SPSS in data analysis.=20
> >> >>=20
> >> >> The names of SPSS procedures often match the names of the
> >> >> tests, such as t-test or ANOVA. =20
> >> >>=20
> >> >> --=20
> >> >> Rich Ulrich
> >> >
> >> >
> >> >
> >> >I tried to conduct a Goodness of fit Chi-Squared but instead of compa=
ring the gender proportions between groups it treated the gender variable i=
ndependently of the groups which is not what I want to do. Am I doing somet=
hing wrong?
> >>=20
> >> Okay.  The procedures often match the names of tests.  On the=20
> >> other hand, there are a lot of occasions where you might=20
> >> apply an F-statistic or a chi-squared, or "goodness-of-fit",=20
> >> and you need an orientation to know which test meets your
> >> circumstances. =20
> >>=20
> >> If you can't take a course, at least read the tutorial.=20
> >>=20
> >> It sounds to me like you wanted a "contingency table" for two
> >> variables but you called upon the "goodness-of-fit" that looked at
> >> each variable alone to see if its cells were all equal.  You can get
> >> surer advice if you show the actual syntax that you ran.  (A good
> >> practice for everyone is to "paste syntax" from the click-and-choose
> >> interface; save your runs for study (at first) and documentation.
> >>=20
> >> --=20
> >> Rich Ulrich
> >
> >
> >I have been reading some of the articles from the UCLA website and I hav=
e to thank you for your suggestion.
> >I do not know how to provide the syntax. So I can just try to give you m=
ore information if that is helpful. Basically I want to see if the three gr=
oups I have differ in gender since I want them to be similar is some variab=
les such as gender, SES, etc. Thus, all I want to verify is if there are di=
fferences between these groups for
> those variables. I treated SES as an ordinal variable as there is an intr=
insic order between the different levels. However, I am struggling to find =
the best way of conducting the differences between gender. I think the most=
 suitable test would be the chi-squared test of independence, but I am not =
100% certain.
>=20
> If you are happy with reporting the average score for
> SES, then SES can be treated as a continuous variable=20
> in an ANOVA.  That is fairly conventional.=20
>=20
> The usual test for a dichotomy like sex would be the
> contingency table.  The test will be almost the same
> p-value as looking at "average sex" in an ANOVA.=20
>=20
> --=20
> Rich Ulrich


I did "crosstabs" and chose pearson's chi-square and Phi and Cramer's V on =
statistics. However, I am still struggling when seeing videos of people usi=
ng this test for relationship between variables when I want to use it for d=
ifferences between groups.
0
snowdenisinrussia
8/6/2016 5:12:52 AM
On Fri, 5 Aug 2016 22:12:52 -0700 (PDT), snowdenisinrussia@gmail.com
wrote:

me>
>> If you are happy with reporting the average score for
>> SES, then SES can be treated as a continuous variable 
>> in an ANOVA.  That is fairly conventional. 
>> 
>> The usual test for a dichotomy like sex would be the
>> contingency table.  The test will be almost the same
>> p-value as looking at "average sex" in an ANOVA. 
>> 

>
>
>I did "crosstabs" and chose pearson's chi-square and Phi and Cramer's V on statistics. However, I am still struggling when seeing videos of people using this test for relationship between variables when I want to use it for differences between groups.

About the time I finished my formal education, I figured out that
if one source did not satisfy me, I could dig out ANOTHER one. 
Sometimes it is not a matter of being "better", but of being more
compatible with what you, personally, already know or assume. 

It might be time to go to a library and browse a few textbooks
used to teach statistics in your field, if you can't find enough
on-line. 

-- 
Rich Ulrich 

0
Rich
8/6/2016 7:05:06 AM
Well,if there is a relationship does that not imply that there is a DIFFERENCE?  Think about it!
0
David
8/11/2016 8:22:08 PM
Reply: