/** Add a new x,y pair. */
public synchronized void addSample(final double x, final double y) {
_xStats.addSample(x);
_yStats.addSample(y);
_xxStats.addSample(x * x);
_xyStats.addSample(x * y);
_yyStats.addSample(y * y);
_needsUpdate = true;
}
/**
* Get the mean square error of the y value with respect to the fitted line. The square root of
* this number gives an indication of the uncertainty in y value estimates under certain
* assumptions about the error distribution.
*
* @return the mean square error of the y value with respect to the fitted line
*/
public synchronized double getMeanSquareOrdinateError() {
performFitIfNeeded();
final double slope = _slope;
final double intercept = _intercept;
final double xyMean = _xyStats.mean();
final double xxMean = _xxStats.mean();
final double yyMean = _yyStats.mean();
return yyMean - 2 * slope * xyMean + slope * slope * xxMean - intercept * intercept;
}
/** Calculate the slope and intercept. */
protected synchronized void performFit() {
final double xMean = _xStats.mean();
final double yMean = _yStats.mean();
final double xyMean = _xyStats.mean();
final double xxMean = _xxStats.mean();
final double yyMean = _yyStats.mean();
_slope = (xyMean - xMean * yMean) / (xxMean - xMean * xMean);
_intercept = yMean - _slope * xMean;
_correlationCoefficient =
(xyMean - xMean * yMean) / Math.sqrt((xxMean - xMean * xMean) * (yyMean - yMean * yMean));
_needsUpdate = false;
}
/**
* Generate a string representation of the linear equation.
*
* @return a string representation of the linear equation
*/
public synchronized String toString() {
performFitIfNeeded();
StringBuffer buffer = new StringBuffer();
buffer.append("y = " + _slope + " * x + " + _intercept);
buffer.append("\n" + "r = " + _correlationCoefficient);
buffer.append("\n" + "<x> = " + _xStats.mean() + ", <y> = " + _yStats.mean());
buffer.append(
"\n"
+ "<xx> = "
+ _xxStats.mean()
+ ", <xy> = "
+ _xyStats.mean()
+ ", <yy> = "
+ _yyStats.mean());
return buffer.toString();
}
