r.kappa: Add MCC

[marisn]

This PR adds Matthews (Mattheus) Correlation Coefficient as one of indices provided by r.kappa.

Should be applied on the top of PR #2666

[marisn]

Ping @wenzeslaus @nilason

[nilason]

Looks good to me!

[wenzeslaus]

The Neumaier's Kahan-Babuska algorithm fits with the version in Wikipedia. I did not check MCC itself.

Given that Kahan-Babuska is applied to existing code too, this may change some results, no? I see it does not change any of the tested code. I think it is the right thing to do and that we can do that between minor versions, but we may need to emphasize that in the release notes (as a separate item in "changed" besides addition to the MCC item).

I'm asking for some code tweaks in the comments.

[marisn]

Given that Kahan-Babuska is applied to existing code too, this may change some results, no? I see it does not change any of the tested code. I think it is the right thing to do and that we can do that between minor versions, but we may need to emphasize that in the release notes (as a separate item in "changed" besides addition to the MCC item).

In most of cases results shouldn't change. NKB will give different outcome only in rare corner cases or when n is large.

[wenzeslaus]

Is there a some kind of style guideline [for using a += b instead of a = a + b]? I do use += for counters and accumulators in loops but prefer to have all terms explicit when summation is neither of them.

While in math it is common to have the terms explicit, the current programming practices usually aim primarily at documenting intention given that people spend most of the time by reading code (theirs or someone else's), so a += b comes out better because it clearly says add b to a.

In the following code:

    p0 = p0 + p0c;
    pC = p0 + pCc;

did I really mean to do p0 + pCc or is it a copy-paste error and I actually mean pC + pCc?

When using +=, it becomes clear:

    p0 += p0c;
    pC += pCc;

Clearly I meant to add to the existing value. If I would really mean p0 + pCc, I would write:

    p0 += p0c;
    pC = p0 + pCc;

[marisn]

did I really mean to do p0 + pCc or is it a copy-paste error and I actually mean pC + pCc?

This is a good reason. Thank you for pointing out to this aspect of coding style.

[wenzeslaus]

You are welcome. I'm sorry I was not more clear before.