"Conditional chis square test of independence"

noetsi

Fortran must die
#1
I am studying for comps (my life is complete):shakehead and I ran into a test that I had not heard of it. Or a variation of one anyhow. It has a conditional chi square test for independence run at each level of one of the two variables (and an associated signficance value).

Any one run into this. I assume if a given level is signficant it supports the overall assumption of dependence, but I never ran into this before.
 

noetsi

Fortran must die
#3
I can't be more clear because I have never actually seen this test and I could not find it on the internet. :( This is the comp question that raised my question:

"In the marginal table [I assume they mean a contingency table with education added although no details are given]... the chi squared value of 35.8 for the independence test is extremely high. On the other hand, for each level of education, the conditional chi squared test for independence of Aptitude and occupation is substantially lower, and three out of four of these tests fail to reach significance."

Aptitude and Occupation are the two variables being tested with education apparently being a control. I do not have the actual data that those taking the test looked at, just the questions.

Looks like they are calculating a seperate chi square for each level of education, but I have never seen chi squared done this way.