Confidence and prediction intervals for uncertainty quantification.

Overview#

Confidence and Prediction Intervals

Adapter support: Confidence and prediction intervals are available in Batch (Lowess) mode only. StreamingLowess and OnlineLowess do not support intervals.

TypeRepresentsWidthUse
ConfidenceUncertainty in mean curveNarrowWhere is the true trend?
PredictionUncertainty for new pointsWideWhere will new data fall?

Confidence Intervals#

Estimate uncertainty in the smoothed curve itself.

import fastlowess.Lowess;
import fastlowess.Options;
import fastlowess.Result;

public class Example {
    public static void main(String[] args) {
        int n = 100;
        double[] x = new double[n];
        double[] y = new double[n];
        for (int i = 0; i < n; i++) {
            x[i] = i * 2 * Math.PI / (n - 1);
            y[i] = Math.sin(x[i]) + 0.1;
        }

        Options options = Options.builder()
                .fraction(0.5)
                .confidenceIntervals(0.95) // 95% CI
                .build();

        try (Lowess model = new Lowess(options)) {
            Result result = model.fit(x, y);

            double[] lower = result.confidenceLower().orElseThrow();
            double[] upper = result.confidenceUpper().orElseThrow();
            for (int i = 0; i < 3; i++) {
                System.out.printf("x=%.2f: y=%.2f [%.2f, %.2f]%n",
                        result.x()[i], result.y()[i], lower[i], upper[i]);
            }
        }
    }
}
x=0.00: y=0.33 [0.29, 0.37]
x=0.06: y=0.36 [0.32, 0.40]
x=0.13: y=0.39 [0.34, 0.43]

Prediction Intervals#

Estimate where new observations might fall.

import fastlowess.Lowess;
import fastlowess.Options;
import fastlowess.Result;

public class Example {
    public static void main(String[] args) {
        int n = 100;
        double[] x = new double[n];
        double[] y = new double[n];
        for (int i = 0; i < n; i++) {
            x[i] = i * 2 * Math.PI / (n - 1);
            y[i] = Math.sin(x[i]) + 0.1;
        }

        Options options = Options.builder()
                .fraction(0.5)
                .predictionIntervals(0.95) // 95% PI
                .build();

        try (Lowess model = new Lowess(options)) {
            Result result = model.fit(x, y);
            System.out.printf("Prediction bounds: [%.2f, %.2f]%n",
                    result.predictionLower().orElseThrow()[0], result.predictionUpper().orElseThrow()[0]);
        }
    }
}
Prediction bounds: [-0.04, 0.71]

Both Intervals#

Request both types simultaneously:

import fastlowess.Lowess;
import fastlowess.Options;
import fastlowess.Result;

public class Example {
    public static void main(String[] args) {
        int n = 100;
        double[] x = new double[n];
        double[] y = new double[n];
        for (int i = 0; i < n; i++) {
            x[i] = i * 2 * Math.PI / (n - 1);
            y[i] = Math.sin(x[i]) + 0.1;
        }

        Options options = Options.builder()
                .fraction(0.5)
                .confidenceIntervals(0.95)
                .predictionIntervals(0.95)
                .build();

        try (Lowess model = new Lowess(options)) {
            Result result = model.fit(x, y);
            System.out.printf("First point 95%% CI: [%s, %s]%n",
                    result.confidenceLower().orElseThrow()[0], result.confidenceUpper().orElseThrow()[0]);
        }
    }
}
First point 95% CI: [0.29412988645250643, 0.3746108343833247]

Confidence Levels#

Common levels and their z-values:

Levelz-valueInterpretation
0.901.64590% of intervals contain true value
0.951.96095% of intervals contain true value
0.992.57699% of intervals contain true value
import fastlowess.Lowess;
import fastlowess.Options;
import fastlowess.Result;

public class Example {
    public static void main(String[] args) {
        int n = 100;
        double[] x = new double[n];
        double[] y = new double[n];
        for (int i = 0; i < n; i++) {
            x[i] = i * 2 * Math.PI / (n - 1);
            y[i] = Math.sin(x[i]) + 0.1;
        }

        // 99% confidence interval
        Options options = Options.builder().confidenceIntervals(0.99).build();

        try (Lowess model = new Lowess(options)) {
            Result result = model.fit(x, y);
            System.out.println("First lower CI bound (99%): " + result.confidenceLower().orElseThrow()[0]);
        }
    }
}
First lower CI bound (99%): 0.31714272639100155

Standard Errors#

Access standard errors directly (available when intervals are computed):

import fastlowess.Lowess;
import fastlowess.Options;
import fastlowess.Result;

public class Example {
    public static void main(String[] args) {
        int n = 100;
        double[] x = new double[n];
        double[] y = new double[n];
        for (int i = 0; i < n; i++) {
            x[i] = i * 2 * Math.PI / (n - 1);
            y[i] = Math.sin(x[i]) + 0.1;
        }

        Options options = Options.builder().returnSe(true).build();

        try (Lowess model = new Lowess(options)) {
            Result result = model.fit(x, y);
            double[] se = result.standardErrors().orElseThrow();
            for (int i = 0; i < 3; i++) {
                System.out.printf("Point %d: SE = %.4f%n", i, se[i]);
            }
        }
    }
}
Point 0: SE = 0.0254
Point 1: SE = 0.0269
Point 2: SE = 0.0283

Availability#

Batch Mode Only: Confidence and prediction intervals are only available in Batch mode. StreamingLowess and OnlineLowess do not support intervals.

FeatureBatchStreamingOnline
Confidence intervals✓✗✗
Prediction intervals✓✗✗
Standard errors✓✗✗

All examples above use the same 100-point noisy sine wave (see Weight Functions for the generation code).