Correlation Structure
Correlation structures are used to model within-group correlation not captured by the random effects. These are generally associated with temporal or spatial dependencies.
Type The type of correlation structure. The available correlation structures are:
None, the default
AR(1), an autoregressive process of order 1, (corAR1)
ARMA(p,q), an autoregressive moving average process, with arbitrary orders for the autoregressive and moving average components (corARMA)
Continuous AR(1), a continuous autoregressive process (that is, an AR(1) process for a continuous time covariate) (corCAR1)
Compound Symmetric, a compound symmetry structure corresponding to a constant correlation (corCompSymm)
General Symmetric, a general correlation matrix, with no additional structure (corSymm)
Exponential, an exponential spatial correlation (corExp)
Gaussian, a Gaussian spatial correlation (corGaus)
Linear, a linear spatial correlation (corLin)
Rational Quadratic, a Rational Quadratic spatial correlation. (corRatio)
Spherical, a spherical spatial correlation (corSpher)
Banded
See the online Help for the corresponding corStruct S-PLUS function to determine the specific parameters and covariates for each of the correlation structures.
Time Variable The continuous time covariate for correlation structures AR(1), ARMA(p,q), Continuous AR(1), Compound Symmetric, and General Symmetric.
Spatial Variables The spatial covariate(s) for correlation structures Exponential, Gaussian, Linear, Rational Quadratic, and Spherical.
Group Variable The grouping factor(s). Multiple grouping factors are converted to ordered factors.
Parameters Parameters as needed for the different correlation structures.