Dear Colleagues:
A co-worker wants to make a correlation matrix of binary variables.
Each variable is a 1/0 indicator of whether a hospital is practicing state-of-art cancer care on various types of cancer.
The data looks like this:
Hospital_ID Compliant_BreastCancer Compliant_Colon Compliant_Lung Compliant_Rectum etc.
A 1 01 1
B 0 1 11
C 1 1 1 0
Does anyone have recommendations for articles or websites that indicate the best measure of binary correlation? Off hand, I think either the phi coefficient or the C statistic from logistic regression would be good options. However, I'm not sure whether the phi coefficient or the C statistic or something else is the best option.
------------------------------
Brandy Sinco, BS, MA, MS
Statistician Senior
Michigan Medicine
------------------------------