Beginner question about correlating subsets

[deca]

Say I have a variable with only two values: people with education level above and below high school.

Then I have other variables lie gender, nationality, race, marital status.

Running crosstabs Chi-square tests on these only gives me the overall value for statistical significance, for example for nationality. What if I wanted to know if any specific nationality or race etc was more likely to have a HS degree or was more likely not to have one? Basically which category of a nominal variable was more likely to have a degree or not have one.

I'm sorry about the strange way my question was worded, but I suspect my problem stems at least partially from the way I'm seeing it at the moment.

Thanks!
Almir

[Bruce Weaver]

Are you asking about how to partition an overall chi-square?  If so, maybe this will help you:

http://www-01.ibm.com/support/docview.wss?uid=swg21501724

deca wrote

Say I have a variable with only two values: people with education level above and below high school.

Then I have other variables lie gender, nationality, race, marital status.

Running crosstabs Chi-square tests on these only gives me the overall value for statistical significance, for example for nationality. What if I wanted to know if any specific nationality or race etc was more likely to have a HS degree or was more likely not to have one? Basically which category of a nominal variable was more likely to have a degree or not have one.

I'm sorry about the strange way my question was worded, but I suspect my problem stems at least partially from the way I'm seeing it at the moment.

Thanks!
Almir

[deca]

Thank you, Bruce. That is pretty much what I want to do, though admittedly I was hoping for a somewhat simpler solution. All I really want to know is which variable subsets are more associated with the dependent variable.

Thanks again for taking the time to reply!

Almir

[Rich Ulrich]

I think that Bruce gives a solution for a bigger problem.

For your question - In order to see which cell gives the biggest
contribution to the chi-squared, you can let SPSS give you the
standardized cell deviations.  Plus or minus signs tell which direction.

--
Rich Ulrich

> Date: Fri, 25 Apr 2014 07:18:11 -0700
> From: [hidden email]
> Subject: Beginner question about correlating subsets
> To: [hidden email]
>
> Say I have a variable with only two values: people with education level above
> and below high school.
>
> Then I have other variables lie gender, nationality, race, marital status.
>
> Running crosstabs Chi-square tests on these only gives me the overall value
> for statistical significance, for example for nationality. What if I wanted
> to know if any specific nationality or race etc was more likely to have a HS
> degree or was more likely not to have one? Basically which category of a
> nominal variable was more likely to have a degree or not have one.
>
> I'm sorry about the strange way my question was worded, but I suspect my
> problem stems at least partially from the way I'm seeing it at the moment.
>

...