public static bool TryGet <TRegression>(IList <double> listIndependent, IList <double> listDependent,
int minPoints, out IIrtRegression regression, IList <Tuple <double, double> > removedValues = null)
where TRegression : IIrtRegression, new()
{
return(TryGet(new TRegression(), listIndependent, listDependent,
minPoints, out regression, removedValues));
}
public static double R(IIrtRegression regression)
{
if (regression?.XValues == null || regression.YValues == null || regression.XValues.Length == 0 || regression.YValues.Length == 0)
{
return(double.NaN);
}
var yMean = new Statistics(regression.YValues).Mean();
var totalSumOfSquares = regression.YValues.Sum(y => (y - yMean) * (y - yMean));
var sumOfSquaresOfResiduals = regression.XValues.Select((x, i) => Math.Pow(regression.YValues[i] - regression.GetY(x), 2)).Sum();
return(Math.Sqrt(1 - sumOfSquaresOfResiduals / totalSumOfSquares));
}
private double GetR(IIrtRegression regression, bool includeOutliers)
{
var r = IrtRegression.R(regression);
if (!double.IsNaN(r))
{
return(r);
}
var pointSet = !includeOutliers?RegularPoints.ToArray() : RegularPoints.Concat(OutlierPoints).ToArray();
var statsX = new Statistics(pointSet.Select(point => point.X));
var statsY = new Statistics(pointSet.Select(point => point.Y));
return(statsY.R(statsX));
}
public static bool TryGet(IIrtRegression existing, IList <double> listIndependent, IList <double> listDependent,
int minPoints, out IIrtRegression regression, IList <Tuple <double, double> > removedValues = null)
{
regression = null;
removedValues?.Clear();
if (listIndependent.Count != listDependent.Count || listIndependent.Count < minPoints)
{
return(false);
}
var listX = new List <double>(listIndependent);
var listY = new List <double>(listDependent);
while (true)
{
regression = existing.ChangePoints(listX.ToArray(), listY.ToArray());
if (Accept(regression, minPoints) || listX.Count <= minPoints)
{
break;
}
var furthest = 0;
var maxDistance = 0.0;
for (var i = 0; i < listY.Count; i++)
{
var distance = Math.Abs(regression.GetY(listX[i]) - listY[i]);
if (distance > maxDistance)
{
furthest = i;
maxDistance = distance;
}
}
removedValues?.Add(new Tuple <double, double>(listX[furthest], listY[furthest]));
listX.RemoveAt(furthest);
listY.RemoveAt(furthest);
}
return(Accept(regression, minPoints));
}
public static void GetCurve(IIrtRegression regression, RetentionTimeStatistics statistics, out double[] hydroScores, out double[] predictions)
{
var minHydro = double.MaxValue;
var maxHydro = double.MinValue;
foreach (var hydroScore in statistics.ListHydroScores)
{
minHydro = Math.Min(minHydro, hydroScore);
maxHydro = Math.Max(maxHydro, hydroScore);
}
var sortedX = regression.XValues.OrderBy(x => x).ToList();
if (minHydro < sortedX[0])
{
sortedX.Insert(0, minHydro);
}
if (maxHydro > sortedX[sortedX.Count - 1])
{
sortedX.Add(maxHydro);
}
hydroScores = sortedX.ToArray();
predictions = sortedX.Select(regression.GetY).ToArray();
}
public static bool Accept(IIrtRegression regression, int minPoints)
{
return(regression.XValues != null && regression.XValues.Length >= minPoints && R(regression) >= RCalcIrt.MIN_IRT_TO_TIME_CORRELATION);
}
