Vito Prosciutto: Teaching community college math on the road to a PhD.

Monday, April 25, 2005

Chi-squared question: Need stats wizards 

Is chi-squared the right approach to this problem or is something else? We want to see if conditions A and B are correlated. We set up a table of data something like this:
Where w, x, y, z are the number of observations for each case. Is chi-squared the right tool to use here? What should my null hypothesis be? Suppose my claim is that there is no correlation? Then I would want expected values of w+x 0 0 y+z, which won't work with the chi-squared formula. Any help here? I'm not really an applied math guy.

This page is powered by Blogger. Isn't yours? Site Meter Listed on Blogwise