/*
* Salichos Stamatakis and Rokas, MBE v31 p1261 (2014)
* If only two splits are listed, returns IC, the Internode Certainty.
* If more splits are listed, returns ICA, (IC All).
*/
private double internodeCertainty(DoubleList<Split, Integer> splits) {
int sum = 0;
int n = splits.size(); // number of splits under consideration
for (int i = 0; i < n; i++) sum += splits.getB(i);
double ic = 1;
if (n > 1) {
ic = Math.log(n);
for (int i = 0; i < n; i++) {
double p = ((double) splits.getB(i)) / sum;
ic += p * Math.log(p);
}
ic /= Math.log(n); // convert from natural log to log base n
}
return ic;
}
/**
* @return f[] where f[i]==c indicates there were c splits which were present in exactly i trees.
*/
public int[] countByFrequency() {
int max = 0;
if (splitsAddedOnlyViaTrees) max = nTrees;
else for (int n : counts.values()) max = Math.max(max, n);
int[] freq = new int[max];
for (int i : counts.values()) freq[i - 1]++;
return freq;
}
/**
* Returns the sum of the Robinson Foulds distances between the majority rules consensus tree and
* each tree in the collection. This does not actually require calculating the majority rule tree.
*
* @return
*/
public int sumRFtoMajRuleTree() {
// Somebody used 'addSplitSystem()' method to supply splits to this count.
if (!splitsAddedOnlyViaTrees)
throw new RuntimeException(
"Can't determine consensus tree unless splits added only via trees");
int sumDist = 0;
for (Integer count : counts.values()) {
sumDist += Math.min(count, nTrees - count);
}
return sumDist;
}