Family Wise Error, Help determining family size...

[Wolphren]

Hello,

I am currently performing odds ratios (with chi squares) on the following table:

Eyes Ears Nasal Oral Respiratory Trauma Gender Age​

Outcome
Gender
Age

I know I need to use a family wise error post hoc test, but I am a bit confused on what is considered a family. Would I consider all the test as a family (thus running 18 tests) or do I just consider the test per row (thus 3 different sets of families)?

Any help would be greatly appreciated.

 

[hlsmith]

Please better explain the test variables and number of test you are running. Are you running 8 tests on outcome, gender, and age. What you doing is unclear?

 

[Wolphren]

Hello, sorry for not making myself clear.

All variables are binomial (e.g. They either have a condition or don't, they are either male or female, outcome is either dead or alive) except age which are ordinal classifications (not human, exact age unknown but can be classified as neonate, yearling, adult). I have run Chi Squares on Outcome Vs Eyes, Outcome Vs Ears, etc. I have also calculated odds ratios on significant tests (e.g. a male in this population is 10 times more likely to die than females, etc). But since I am running so many test, I need to account for the family wise error. I am just confused on what is considered a family (e.g. do I run a Sidak-Bonferroni with C=18, or treat each row as a separate family and have C=8).

Once again, thanks for any help!

 

[Wolphren]

Hi, I was hoping someone had an answer I could use.

Thanks.