chi square feature analysis

it attempts to analyze the relevance between features and labels, for feature selection, etc.

parameter 1 is feature value (discrete), parameter 2 is the corresponding labels

the reference paper is: S. Wu, P. A. Flach, Feature selection with labeled and unlabeled data, available at
http://ai.ijs.si/branax/iddm-2002_proceedings/Wu.pdf

relevant paper is: Technical Note: Bias in Information-Based Measures in Decision Tree Induction.

I programmed the file according to the first paper, but I find that the results are not reasonable, wondering if it is the drawback of chi square feature analysis method or some bugs in my file. Hope somebody can point out.

The computed chi value needs to compared with chi square table to see how important are the features.
chi square table:
http://www.ma.utexas.edu/users/davis/375/popecol/tables/chisq.html