reStruct-class {lme4} | R Documentation |
The random-effects model structure in a linear mixed-effects model or a generalized linear mixed-effects model or a nonlinear mixed-effects model.
Objects can be created by calls of the form new("reStruct", ...)
.
random
:"list"
of lmeLevel
objects
giving the levels of random effects in the model, the
fixed-effects, and the response or the working residual.fixed
:"formula"
of the response and
fixed-effects terms.offset
:"numeric"
: the offset in the model, if present.dirtyDecomposed
:"logical"
: if TRUE
the
decomposed matrix must be recalculated.useWeighted
:"logical"
: if TRUE
calculate and use the weighted model matrix to form the
decompositions; otherwise use the original model matrix.dirtyStored
:"logical"
: if TRUE
the
matrix of stored decompositions must be recalculated.dirtyBbetas
:"logical"
: if TRUE
the
BLUP's and the conditional estimates of the fixed-effects must be
recalculated.logLik
:"numeric"
: the log-likelihood at the
current parameter values.REML
:"logical"
: if TRUE
the parameters
will be estimated according to the REML criterionanalyticHessian
:"logical"
: if TRUE
the
hessianArray
slot of the lmeLevel objects are calculated
during optimization.reverseOrder
:"integer"
: the permutation of the
rows that provides the original order.origOrder
:"integer"
: the inverse of the
reverseOrder
permutation.original
:"matrix"
: the original model matrix,
including the original response in the last column.weighted
:"matrix"
: the model matrix and
responses after applying weights.stored
:"matrix"
: a model matrix of
intermediate decomposition results needed for evaluating the BLUPs
and the EM or ECME iterations.decomposed
:"matrix"
: the model matrix after
predecomposition. This generally has many few rows than
original
.bbetas
:"numeric"
: the BLUPs and the conditional
estimates of the fixed-effects parameters at the current values
of the relative precision matrices.dontCopy
:"logical"
: if TRUE
it indicates
that this object has just been created and is only assigned to one
name. In these circumstances changes are made directly on the
object without copying. This is dangerous. You probably don't
want to modify this setting.assign.X
:"ANY"
: the
assign
attribute from the model matrix for the fixed effects.signature(x = "reStruct", value = "list")
:
perform the EM iterationssignature(x = "reStruct", A = "missing",
nlev = "missing")
: evaluate the gradient of the linear
mixed-effects profiled log-likelihood.signature(x = "reStruct", value =
"list")
: Optimize with optim
or nlm
.signature(x = "reStruct")
: Extract the
variances and covariances of the random effects.signature(object = "reStruct")
: return the
current parameters in the object.signature(object = "reStruct", value =
"numeric")
: assign the parameters to the object.signature(object = "reStruct")
: extract the
fitted values.signature(object = "reStruct")
: extract the
fixed effects.signature(object = "reStruct", value =
"numeric")
: assign the fixed effects.signature(object = "reStruct")
: extract the
denominator degrees of freedom for the fixed-effects terms.signature(object = "reStruct", form =
"missing", level = "ANY", data = "missing", sep = "missing")
:
return the grouping factors for the random effectssignature(object = "reStruct")
: extract
the response.signature(object = "reStruct")
: return the
log-likelihood at the current parameter values. signature(object = "reStruct")
: return the model
matrix for the object.signature(x = "reStruct", value =
"matrix")
: assign the model matrix to the object.signature(object = "reStruct")
: extract the
random effects predictors.signature(x = "reStruct", value =
"numeric")
: set the working response (used in GLMM models).signature(object = "reStruct")
: summarize the object.signature(x = "reStruct", value =
"matrix")
: update the weights.