Re: Correlating binary variables
- To: mathgroup at smc.vnet.net
- Subject: [mg70302] Re: Correlating binary variables
- From: "Ray Koopman" <koopman at sfu.ca>
- Date: Thu, 12 Oct 2006 05:37:30 -0400 (EDT)
- References: <egi17o$j42$1@smc.vnet.net>
mickey wrote: > Hi, > > I have a problem where I am trying to find the correlation between a > series of binary variables. > > 10101000101010010 > 11010010100101001 > 01010010100100101 > > and so on. What is good way of going about this? Linear regression > doesnt seem to be that good a way of doing things when both variables > are binary. > > Thanks, > -M That argument has been going on for over a hundred years, from the time of Pearson and Yule. (For a brief summary, see chap 10 in Alan Agresti, An Introduction to Categorical Data Analysis, Wiley, 1996.) How you should measure correlation will depend on how you got the data, the probability model that underlies it, and the information you want the correlation to convey.