Variance Models

Variance model keywords

keywords are listed in three groups
  • Correlation matrices
  • Variance matrices
  • Fixed matrices
  • Correlation matrices can be converted to variance matrices.

    The algebraic definitions of these structures are available in the User Guide.

    Correlation models.

    In the following table, w is the order of the matrix and the last column specifies the number of parameter values (initial values expected).
    Keyword Description Parameters
    Correlation matrices
    ID Identity 0
    AR1 First order autoregressive 1
    AR2 Autoregressive 2
    AR3 Autoregressive 3
    SAR Symmetric autoregressive 1
    SAR2 Symmetric autoregressive 2
    MA, MA1 Moving Average 1
    MA2 Moving Average 2
    ARMA Autoregressive Moving Average 2
    CORU Uniform Correlation 1
    CORB Banded Correlation w-1
    CORG General Correlation w(w-1)/2
    EXP Exponential 1
    GAU Gaussian 1
    IEXP Isotropic Exponential 1
    IGAU Isotropic Gaussian 1
    IEUC Isotrophic Euclidean 1
    LVR Linear Variance 1
    SPH Spherical 1
    CIR Circular 1
    AEXP Anisotropic Exponential 2
    AGAU Anisotropic Gaussian 2
    MATk Matern k
    Variance structures
    DIAG Diagonal w
    US Unstructured w(w+1)/2
    OWNk User supplied matrix k
    ANTEk Antedependence w(w+1)/2
    CHOLk Cholesky - banded form w(w+1)/2
    CHOLkC Cholesky - column form w(w+1)/2
    FAk Factor Analytic (correlation form) w(k+1)
    FACV Factor Analytic (covariance form) w(k+1)
    XFA k Extended Factor Analytic w(k+1)
    Fixed matrices
    AINV Numerator Relationship matrix (A) 0 or 1
    GIVk User supplied General (Inverse) Variance matrix 0 or 1

    See Also