Thursday, January 26, 2012

MIC

At last night's Data Science meetup in DC we heard about the Maximal Information Coefficient led by Sean Murphy, recently developed by two brothers. 



It's a non-parametric statistic which helps determine meaningful correlations between a slew of variables, but with out assuming a linear relationship as a correlation matrix or principal components/factor analysis methods do. 

It's useful when you're trying to look for first swipes at data. 

No comments:

Post a Comment