@stdlib/stats-lowess
    TypeScript icon, indicating that this package has built-in type declarations

    0.0.6 • Public • Published

    LOWESS

    NPM version Build Status Coverage Status

    Locally-weighted polynomial regression via the LOWESS algorithm.

    Installation

    npm install @stdlib/stats-lowess

    Usage

    var lowess = require( '@stdlib/stats-lowess' );

    lowess( x, y[, opts] )

    For input arrays and/or typed arrays x and y, the function returns an object holding the ordered input values x and smoothed values for y.

    var x = [
        4, 4, 7, 7, 8, 9, 10, 10, 10, 11, 11, 12, 12, 12, 12, 13, 13, 13, 13, 14,
        14, 14, 14, 15, 15, 15, 16, 16, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 20,
        20, 20, 20, 20, 22, 23, 24, 24, 24, 24, 25
    ];
    var y = [
        2, 10, 4, 22, 16, 10, 18, 26, 34, 17, 28, 14, 20, 24, 28, 26, 34, 34, 46,
        26, 36, 60, 80, 20, 26, 54, 32, 40, 32, 40, 50, 42, 56, 76, 84, 36, 46, 68,
        32, 48, 52, 56, 64, 66, 54, 70, 92, 93, 120, 85
    ];
    
    var out = lowess( x, y );
    /* returns
        {
            'x': [
                4,
                4,
                7,
                7,
                ...,
                24,
                24,
                24,
                25
            ],
            'y': [
                ~4.857,
                ~4.857,
                ~13.1037,
                ~13.1037,
                ...,
                ~79.102,
                ~79.102,
                ~79.102,
                ~84.825
            ]
        }
    */

    The function accepts the following options:

    • f: positive number specifying the smoothing span, i.e., the proportion of points which influence smoothing at each value. Larger values correspond to more smoothing. Default: 2/3.
    • nsteps: number of iterations in the robust fit (fewer iterations translates to faster function execution). If set to zero, the nonrobust fit is returned. Default: 3.
    • delta: nonnegative number which may be used to reduce the number of computations. Default: 1/100th of the range of x.
    • sorted: boolean indicating if the input array x is sorted. Default: false.

    By default, smoothing at each value is determined by 2/3 of all other points. To choose a different smoothing span, set the f option.

    var x = [
        4, 4, 7, 7, 8, 9, 10, 10, 10, 11, 11, 12, 12, 12, 12, 13, 13, 13, 13, 14,
        14, 14, 14, 15, 15, 15, 16, 16, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 20,
        20, 20, 20, 20, 22, 23, 24, 24, 24, 24, 25
    ];
    var y = [
        2, 10, 4, 22, 16, 10, 18, 26, 34, 17, 28, 14, 20, 24, 28, 26, 34, 34, 46,
        26, 36, 60, 80, 20, 26, 54, 32, 40, 32, 40, 50, 42, 56, 76, 84, 36, 46, 68,
        32, 48, 52, 56, 64, 66, 54, 70, 92, 93, 120, 85
    ];
    
    var out = lowess( x, y, {
        'f': 0.2
    });
    /* returns
        {
            'x': [
                4,
                4,
                7,
                ...,
                24,
                24,
                25
            ],
            'y': [
                ~6.03,
                ~6.03,
                ~12.68,
                ...,
                ~82.575,
                ~82.575,
                ~93.028
            ]
        }
    */

    By default, three iterations of locally weighted regression fits are calculated after the initial fit. To set a different number of iterations for the robust fit, set the nsteps option.

    var x = [
        4, 4, 7, 7, 8, 9, 10, 10, 10, 11, 11, 12, 12, 12, 12, 13, 13, 13, 13, 14,
        14, 14, 14, 15, 15, 15, 16, 16, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 20,
        20, 20, 20, 20, 22, 23, 24, 24, 24, 24, 25
    ];
    var y = [
        2, 10, 4, 22, 16, 10, 18, 26, 34, 17, 28, 14, 20, 24, 28, 26, 34, 34, 46,
        26, 36, 60, 80, 20, 26, 54, 32, 40, 32, 40, 50, 42, 56, 76, 84, 36, 46, 68,
        32, 48, 52, 56, 64, 66, 54, 70, 92, 93, 120, 85
    ];
    
    var out = lowess( x, y, {
        'nsteps': 20
    });
    /* returns
        {
            'x': [
                4,
                ...,
                25
            ],
            'y': [
                ~4.857,
                ...,
                ~84.825
            ]
        }
    */

    To save computations, set the delta option. For cases where one has a large number of (x,y)-pairs, carrying out regression calculations for all points is not likely to be necessary. By default, delta is set to 1/100th of the range of the values in x. In this case, if the values in x were uniformly scattered over the entire range, the locally weighted regression would be calculated at approximately 100 points. On the other hand, for small data sets with less than 100 observations, one can safely set the option to zero so calculations are made for each data point.

    var x = [
        4, 4, 7, 7, 8, 9, 10, 10, 10, 11, 11, 12, 12, 12, 12, 13, 13, 13, 13, 14,
        14, 14, 14, 15, 15, 15, 16, 16, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 20,
        20, 20, 20, 20, 22, 23, 24, 24, 24, 24, 25
    ];
    var y = [
        2, 10, 4, 22, 16, 10, 18, 26, 34, 17, 28, 14, 20, 24, 28, 26, 34, 34, 46,
        26, 36, 60, 80, 20, 26, 54, 32, 40, 32, 40, 50, 42, 56, 76, 84, 36, 46, 68,
        32, 48, 52, 56, 64, 66, 54, 70, 92, 93, 120, 85
    ];
    
    var out = lowess( x, y, {
        'delta': 0.0
    });
    /* returns
        {
            'x': [
                4,
                ...,
                25
            ],
            'y': [
                ~4.857,
                ...,
                ~84.825
            ]
        }
    */

    If the elements of x are sorted in ascending order, set the sorted option to true.

    var x = [
        4, 4, 7, 7, 8, 9, 10, 10, 10, 11, 11, 12, 12, 12, 12, 13, 13, 13, 13, 14,
        14, 14, 14, 15, 15, 15, 16, 16, 17, 17, 17, 18, 18, 18, 18, 19, 19, 19, 20,
        20, 20, 20, 20, 22, 23, 24, 24, 24, 24, 25
    ];
    var y = [
        2, 10, 4, 22, 16, 10, 18, 26, 34, 17, 28, 14, 20, 24, 28, 26, 34, 34, 46,
        26, 36, 60, 80, 20, 26, 54, 32, 40, 32, 40, 50, 42, 56, 76, 84, 36, 46, 68,
        32, 48, 52, 56, 64, 66, 54, 70, 92, 93, 120, 85
    ];
    
    var out = lowess( x, y, {
        'sorted': true
    });
    /* returns
        {
            'x': [
                4,
                ...,
                25
            ],
            'y': [
                ~4.857,
                ...,
                ~84.825
            ]
        }
    */

    Examples

    var randn = require( '@stdlib/random-base-randn' );
    var Float64Array = require( '@stdlib/array-float64' );
    var plot = require( '@stdlib/plot-ctor' );
    var lowess = require( '@stdlib/stats-lowess' );
    
    var x;
    var y;
    var i;
    
    // Create some data...
    x = new Float64Array( 100 );
    y = new Float64Array( x.length );
    for ( i = 0; i < x.length; i++ ) {
        x[ i ] = i;
        y[ i ] = ( 0.5*i ) + ( 10.0*randn() );
    }
    
    var opts = {
        'delta': 0
    };
    
    var out = lowess( x, y, opts );
    var h = plot( [ x, out.x ], [ y, out.y ] );
    
    h.lineStyle = [ 'none', '-' ];
    h.symbols = [ 'closed-circle', 'none' ];
    
    h.view( 'stdout' );

    References

    • Cleveland, William S. 1979. "Robust Locally and Smoothing Weighted Regression Scatterplots." Journal of the American Statistical Association 74 (368): 829–36. doi:10.1080/01621459.1979.10481038.
    • Cleveland, William S. 1981. "Lowess: A program for smoothing scatterplots by robust locally weighted regression." American Statistician 35 (1): 54–55. doi:10.2307/2683591.

    Notice

    This package is part of stdlib, a standard library for JavaScript and Node.js, with an emphasis on numerical and scientific computing. The library provides a collection of robust, high performance libraries for mathematics, statistics, streams, utilities, and more.

    For more information on the project, filing bug reports and feature requests, and guidance on how to develop stdlib, see the main project repository.

    Community

    Chat


    License

    See LICENSE.

    Copyright

    Copyright © 2016-2022. The Stdlib Authors.

    Install

    npm i @stdlib/stats-lowess

    Homepage

    stdlib.io

    DownloadsWeekly Downloads

    133

    Version

    0.0.6

    License

    Apache-2.0

    Unpacked Size

    65.2 kB

    Total Files

    12

    Last publish

    Collaborators

    • stdlib-bot
    • kgryte
    • planeshifter
    • rreusser