public void add(double yreal, double ymodel) {
null_deviance += _glm.deviance(yreal, _ymu);
if (_glm.family == Family.binomial) // classification -> update confusion matrix too
for (int i = 0; i < DEFAULT_THRESHOLDS.length; ++i)
_cms[i].add((int) yreal, (ymodel >= DEFAULT_THRESHOLDS[i]) ? 1 : 0);
if (Double.isNaN(_glm.deviance(yreal, ymodel)))
System.out.println("NaN from yreal=" + yreal + ", ymodel=" + ymodel);
residual_deviance += _glm.deviance(yreal, ymodel);
++nobs;
avg_err += (ymodel - yreal) * (ymodel - yreal);
if (_glm.family == Family.poisson) { // aic for poisson
long y = Math.round(yreal);
double logfactorial = 0;
for (long i = 2; i <= y; ++i) logfactorial += Math.log(i);
_aic2 += (yreal * Math.log(ymodel) - logfactorial - ymodel);
}
}
