/**
* Estimates the mean of array <code>y</code>. <code>robustMu</code> uses a Gaussian model to
* remove outliers iteratively.
*
* @param y an array
* @return the estimated mean of <code>y</code>
*/
private double robustMu(double[] y) {
ArrayList<Double> x = new ArrayList<Double>(y.length);
for (int i = 0; i < y.length; i++) {
x.add(y[i]);
}
double sigma = 0;
NormalDist gaussian = new NormalDist();
double cutoff = 0;
double cutoff1 = 0;
double cutoff2 = 0;
int xSize = x.size();
sigma = std(x);
double areaRemoved = 0.5 / xSize;
cutoff = abs(gaussian.inverseF(areaRemoved));
cutoff1 = -cutoff * sigma + mean(x);
cutoff2 = cutoff * sigma + mean(x);
while (true) {
Iterator<Double> iter = x.iterator();
int numRemoved = 0;
while (iter.hasNext()) {
double cn = iter.next();
if (cn < cutoff1 || cn > cutoff2) {
numRemoved++;
iter.remove();
}
}
if (numRemoved == 0) {
break;
}
sigma = std(x);
sigma =
sigma
/ sqrt(
(1 - 2 * areaRemoved - 2 * cutoff * exp(-pow(cutoff, 2) / 2) / sqrt(2 * PI))
/ (1 - 2 * areaRemoved));
xSize = x.size();
areaRemoved = 0.5 / xSize;
cutoff = abs(gaussian.inverseF(areaRemoved));
cutoff1 = -cutoff * sigma + mean(x);
cutoff2 = cutoff * sigma + mean(x);
}
return mean(x);
}
