# Chi-square post-hoc test - Helpless

#### alexstats

##### New Member
I have been looking around on the net looking for the best way to do a post-hoc analysis for multi factor significant chi-square test, with no success.
I have a data-set of patients with different disease subtypes (Lung cancer subtype A-H).
So my two main variables are
- Ethnicity (12 levels)
- Disease type (8 levels)

I have performed a chi-square which turned out significant. However I dont know how to answer my following two questions:

1) Which disease subtype is most prevalent in each particular ethnic group
2) which ethnicity is most prevalent in each particular disease subtype.

Here is an example of how my data is structured in SPSS:

Here is an example of my contingency table

Any help is hugely appreciated <3

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Well if you just want prevalences use your contingency table.
A 15/86
German 6/16
...
polish is a 4-way tie.

#### alexstats

##### New Member
Thanks for your reply! but how do I know if
in A, Finish (15/86) is significantly different/greater than American (9/86)?
or if prevalence of A(6/16) is significantly more than B(4/16) in Germans?
I need to show these differences with p-values :-/

#### Karabiner

##### TS Contributor
In your example, you want to perform 66 pairwise comparisons between nations, for each subtype, i.e. a total of 8*66=528 comparisons. And you want to perform 28 comparisons between subtypes, for each nation, i.a. a total of 12*28=336 comparisons. And your 864 pairwise comparisons will be carried out,on just 177 observations. This looks a bit imbalanced.

By the way, a Chi² test for the whole table is not possible here, since about 80-90% of the cells have expected (!) frequencies < 5, most of them even have expected frequencies < 2 or even< 1. The Chi² value is extremely distorted here.

With kind regards

Karabiner