Computing Information Content in Biology

[Patrick]

@EricMH

If I’m right, and ILZ(X:Y)=ILZ(E(X):Y)=0I_{LZ}(X:Y) = I_{LZ}(E(X):Y) = 0 in general, then on average ILZ(X:Y)>ILZ(E(X):Y)I_{LZ}(X:Y) > I_{LZ}(E(X):Y) should fail half the time.

The other control is when X=YX=Y. In this case, ILZ(X:Y)>ILZ(E(X):Y)I_{LZ}(X:Y) > I_{LZ}(E(X):Y) should fail less than half the time.

I don’t agree with this. it seems wrong on the math.