func (o *OLS) String() string {
q, _ := govector.AsVector(o.residuals)
points := []float64{0.0, 0.25, 0.5, 0.75, 1.0}
p, _ := govector.AsVector(points)
qnt := q.Quantiles(p)
f, fp := o.F_Statistic()
return fmt.Sprintf(`
Residuals:
Min 25 50t 75 Max:
%v
Coefficients:
%v
RSS: %v
MSE: %v
Adjusted R-Squared: %v
R-squared: %v
F-statistic: %v with P-value: %v`,
roundAll(qnt),
roundAll(o.betas),
round(o.ResidualSumofSquares(), 3),
round(o.MeanSquaredError(), 3),
round(o.AdjustedRSquared(), 3),
round(o.RSquared(), 3),
round(f, 4), round(fp, 10),
)
}
func (o *OLS) TotalSumofSquares() float64 {
// no chance this could error
y, _ := govector.AsVector(o.response)
ybar := mean(o.response)
squaredDiff := func(x float64) float64 {
return math.Pow(x-ybar, 2.0)
}
return y.Apply(squaredDiff).Sum()
}
// Durbin Watson Test for Autocorrelatoin of the Residuals
// d = \sum_i=2 ^ n (e_i - e_i-1)^2 / \sum_i=1^n e_i^2
//
// Does not calculate the p-value
func (o *OLS) DW_Test() float64 {
e, err := govector.AsVector(o.residuals)
if err != nil {
panic(err)
}
square := func(x float64) float64 { return math.Pow(x, 2) }
d := e.Diff().Apply(square).Sum()
d /= e.Apply(square).Sum()
return d
}
func NewAnomalyzer(conf *AnomalyzerConf, data []float64) (Anomalyzer, error) {
err := validateConf(conf)
if err != nil {
return Anomalyzer{}, err
}
vector, err := govector.AsVector(data)
if err != nil {
return Anomalyzer{}, err
}
return Anomalyzer{conf, vector}, nil
}