mdpd

mdpd computes Minimum Distance Power Divergence statistics

Syntax

Description

The Power Divergence for a density function $f$ and observations $y_1 , \ldots , y_n$ is defined as \[ \Delta(f,\alpha) = \int_R f^{1+\alpha}(y) dy - (1+1/\alpha) \sum_{i=1}^n f^\alpha(y_i)/n \] for $\alpha > 0$ \[ \Delta(f,0) = -\sum_{i=1}^n \log f(y_i)/n \] for $\alpha = 0$.

example

res =mdpd(y, alpha) mdpd with all default arguments.

example

res =mdpd(y, alpha, Name, Value) Student T with 5 degrees of freedom.

Examples

expand all

  • mdpd with all default arguments.
  • rng('default')
    y =randn(10,1);
    out=mdpd(y,0.2);

  • Student T with 5 degrees of freedom.
  • rng('default')
    y =randn(10,1);
    mdpd(y,0.1,'densfunc',@tpdf,'theta',5)
    ans =
    
       -8.7846
    
    

    Related Examples

    expand all

  • Change lower and upper integration limits.
  • rng('default')
    % Generate 10 numbers from Uniform.
    y =mtR(10,0);
    out=mdpd(y,0.1,'densfunc',@tpdf,'theta',5, 'lower',-Inf,'upper',Inf);
    expectedRes = -8.870156744130417;
    assert(isequal(round(out,7), round(expectedRes,7)), 'Error: MATLAB did not output the expected result!')

  • Change lower and upper integration limits and relatve tolerance.
  • Change lower and upper integration limits.

    rng('default')
    % Generate 10 numbers from Uniform.
    y =mtR(10,0);
    out=mdpd(y,0.1,'densfunc',@tpdf,'theta',5, 'lower',-Inf,'upper',Inf,'RelTol',1e-15);
    expectedRes =  -8.870156744130275;
    assert(isequal(round(out,7), round(expectedRes,7)), 'Error: MATLAB did not output the expected result!')

    Input Arguments

    expand all

    y — Response variable. Vector.

    A vector with n elements that contains the response variable. It can be either a row or a column vector.

    Data Types: double

    alpha — Numeric for the power divergence parameter. Non negative scalar.

    It can be shown that as the tuning parameter $\alpha$ increases the robustness of the Minimum Density Power Divergence estimator increases while its efficiency decreases (Basu et al., 1998).

    Data Types: double

    Name-Value Pair Arguments

    Specify optional comma-separated pairs of Name,Value arguments. Name is the argument name and Value is the corresponding value. Name must appear inside single quotes (' '). You can specify several name and value pair arguments in any order as Name1,Value1,...,NameN,ValueN.

    Example: 'densfunc', @tpdf , 'lower',0 , 'upper',10 , 'theta', [10 100] , 'RelTol', 1e-12 , 'AbsTol', 1e-9

    densfunc —handle to the function computing the theoretical density.function handle.

    Function handle which defines the function to be integrated from lower to upper. The default density function is the standard normal distribution.

    Example: 'densfunc', @tpdf

    Data Types: handle

    lower —Lower bound of the domain of the density function.scalar.

    The default value of lower is 1.

    Example: 'lower',0

    Data Types: double

    upper —Upper bound of the domain of the density function.scalar.

    The default value of upper is Inf.

    Example: 'upper',10

    Data Types: double

    theta —The parameters of the distribution given as a vector.numeric vector.

    The default values of theta is [0 1] given that the default density function is the standard normal distribution.

    Example: 'theta', [10 100]

    Data Types: double

    RelTol —Relative error tolerance.non negative scalar.

    Relative error tolerance, specified as the comma-separated pair consisting of 'RelTol' and a nonnegative real number.

    mpdm uses the relative error tolerance to limit an estimate of the relative error, |q - Q| / min(|q|,|Q|), where q is the computed value of the integral and Q is the (unknown) exact value.

    The default value of RelTol is 1e-6.

    Example: 'RelTol', 1e-12

    Data Types: double

    AbsTol —Absolute error tolernace.non negative scalar.

    Absolute error tolerance, specified as the comma-separated pair consisting of 'AbsTol' and a nonnegative real number.

    mpdm uses the absolute error tolerance to limit an estimate of the absolute error, |q - Q|, where q is the computed value of the integral and Q is the (unknown) exact value.

    This can be useful when q or Q becomes close to zero and the relative tolerance risks to go to infinity.

    The default value of AbsTol is 1e-12.

    Example: 'AbsTol', 1e-9

    Data Types: double

    Output Arguments

    expand all

    res —power divergence against the density function densfunc. Scalar

    Value of power divergence wrt the density specified in densfunc.

    Data Types - double.

    References

    Basu, A., Harris, I.R., Hjort, N.L. and Jones, M.C., (1998), Robust and efficient estimation by minimizing a density power divergence, Biometrika, 85, pp. 549-559.

    Riani, M. Atkinson, A.C., Corbellini A. and Perrotta A. (2020), Robust Regression with Density Power Divergence: Theory, Comparisons and Data Analysis, Entropy, Vol. 22, 399.

    https://www.mdpi.com/1099-4300/22/4/399

    This page has been automatically generated by our routine publishFS