This (S3) class object contains a fitted linear mixed model from the asreml() function

Description

Objects of this class have methods for the generic functions wald(), coef(), fitted(), plot(), predict(), resid(), summary() and update().

Value

A list object with class asreml; the following components are included in a valid asreml object:

• loglik: The log-likelihood at completion of the asreml call.

• vparameters: The vector of variance components estimates from the model fit.

• vparameters.con: A numeric vector identifying the boundary constraint applied to each variance parameter at termination. Common values are 1, 3 and 4 for Positive, Unconstrained and Fixed, respectively.

• vparameters.type: A numeric vector identifying the variance parameter types.

• vparameters.pc: Percentage change in gammas parameters over the last iteration.

• score: The score vector of length number of random parameters.

• coefficients: A list with three components named fixed, random and sparse containing the solutions to the mixed model equations corresponding to the E-BLUEs of the fixed effects, the E-BLUPs of the random effects, and the solutions corresponding to the sparse fixed effects, respectively. The coefficients are labelled by a concatenation of factor name and level separated by "_".

• vcoeff: A list with three components named fixed, random and sparse containing the unscaled variances of the coefficients. The actual variances are calculated as vcoeff*object\$sigma2 and returned by the summary function.

• predictions: If predict is not NULL, a list object with components pvals, sed, vcov and avsed. The predictions component only is returned by the predict method for asreml objects.

• fitted.values: A vector containing the fitted values from the model, obtained by transforming the linear predictors by the inverse of the link function.

• linear.predictors: The linear fit on the link scale.

• residuals: A single-column matrix containing the residuals from the model.

• hat: The diagonal elements of the matrix W C^{-1} W^T, the extended hat matrix. This is the linear mixed effects model analogue of X(X^T X)^{-1} X^T for ordinary linear models.

• sigma2: The REML estimate of the scale parameter.

• deviance: The deviance from the fit.

• nedf: The residual degrees of freedom, length(y) - rank(X).

• nwv: The number of working variables.

• noeff: A vector containing the number of effects for each term.

• yssqu: A vector of incremental sums of squares for (dense) fixed terms.

• ai: The inverse average information matrix of the variance parameters. A Matrix class object, sub-class dspMatrix is returned.

• Cfixed: Reflexive generalized inverse of the coefficient matrix of the mixed model equations relating to the dense fixed effects (if asreml.options()$Cfixed = TRUE, a matrix of class Matrix, sub-class dspMatrix is returned).

• Csparse: If asreml.options()$Csparse is not NULL, the non-zero elements of the reflexive generalized inverse matrix of the coefficient matrix for the sparse stored model terms nominated in the Csparse formula. A matrix in triplet form giving the row, column and non-zero elements.

• design: The design matrix as a sparse Matrix of class dgCMatrix if asreml.options(design = TRUE).

• call: An image of the asreml function call.

• trace: A numeric matrix recording the convergence sequence for each random component, as well as the log-likelihood, residual variance and residual degrees of freedom.

• license: A character string containing the license information. The string has embedded new-line characters and is best formatted through cat().

• G.param: A list object containing the constraints and final estimates of the variance parameters relating to the random part of the model. This object may be used as the value of the G.param argument to provide initial parameter estimates to asreml.

• R.param: A list object containing the constraints and final estimates of the variance parameters relating to the error structure of the model. This object may be used as the value of the R.param argument to provide initial parameter estimates to asreml.

• formulae: A list object containing the fixed, random, sparse and residual formula arguments to asreml.

• factor.names: A character vector of term names appearing in the model.

• meff: Regressor scores (marker effects) if nominated in the mef list argument.

• mf: The model frame with the data as a data.table object with numerous attributes from the model specification. Inspect names(attr(object$mf)) for details.

• wrkv: The residual working variables (if requested and available).