BackMIC

A new algorithm for the calculation of the Maximal Information Coeffecient(MIC)

BEFORE STARTING

Make sure that c++ has installed in your computer;

Make sure Matlab has installed in your computer.

1. Down load all the files:BackMIC.m, equipartitionYaxis2c.c,getmutualI2var_fix4.c,getsuper2var.c and make_mex.m.

2. run“make_mex.m” to compile the equipartitionYaxis2c.c, getsuper2var.c and getmutualI2var_fix4.c to mex files;

3. Preparing a paired variable X and Y to measure their correlation. X and Y are both column vectors with the same sample size n.

4. Before computing the MIC of X and Y with BackMIC, you need first scramble the samples, the codes are as follows:

data=[X,Y];

num=randperm(size(data,1));

data=data(num',:);

5. Run "BackMIC.m”to obtain the MIC value of X and Y.

[BackMIC,thebestc1,thebestc2]=BackMIC(data,B(n),c,n,threshold );

n is the sample size;

B(n) is always set to n^0.6;

c is always set to 5;

threshold is always set to 0.01.

thebestc1 and thebestc2 are the segment point along the x-axis and y-axis.