Chi square test

Ste90

New Member
#1
I decided to use Chi quare test because I have a Indipendent categorical variable (with 2 levels) and a dependent categorical variable (with 5 levels).
I ran Descriptives, Crosstabs, I selected Chi square, observed and expected results, row percentages an in Nominal Phi and Cramer's v.
I got significance because I got a Pearson coefficient 13.808a and a p-value = .017 / Likelihood Ratio 18.312 and p value = .003 (df=5)
But, I violated the assumption because reported a 5 cells (41.7%) have expected count less than 5. The minimum expected count is .89.

What should I do now? Did I do something wrong? Is my relation significant even if I violated the assumption? Thanks a lot.
 
Last edited:

Karabiner

TS Contributor
#2
Either your sample size is quite small, or half of the categories in your variable are not very relevant.
What is your dependent variable with 5 levels, and how large is your sample size?

Since you are asking this in SPSS, maybe you have installed "exact tests"?

With kind regards

Karabiner
 

Ste90

New Member
#3
My depent variable is with 6 levels actually, pardon, my sample 114
Since you are asking this in SPSS, maybe you have installed "exact tests"? Where can I see this option?
Thanks a lot
 

Ste90

New Member
#4
I found "exact tests" option. When I ran the chi square I selected Asymptotic only. Is it correct? Other two options are "Monte Carlo" and "Exact"
 

Karabiner

TS Contributor
#5
Is it really important and interpretable to compare those small groups with regard to
a categorical variabe with 6 levels? Otherwise, you could maybe create broader
categories by collapsing some of the current categories.

As far as I know, in the SPSS menue for crosstabs there appears a button "Exact" if the
exact tests module was installed.

With kind regards

Karabiner
 

Ste90

New Member
#6
I see your point. you are suggesting of reducing the 6 variables. I don't know If I can, I will think about it.
About the button exact, it appears . Is my choice correct?

exact tests.PNG
 

Ste90

New Member
#7
Btw you were right Karabiner! If I create broader variable the assumption violated is reduced, now is 30% I try to reach the 20
 

gianmarco

TS Contributor
#8
Just a supplemetary note:
you may want to inspect the (adjusted) standardized residuals to find out which cell significantly contribute to the departure from independence.

Best
Gm
 

gianmarco

TS Contributor
#10
Hello,
i cant remember but there should be an option in SPSS, either something to tick or a button to press (that would lead you to another menu).
Apologies if I cant be more specific, but it's a while that i dont use SPSS (which I do not particularly like by the way).

Gm