R: Model term constructor functions

asr_dsum {asreml}

R Documentation

Model term constructor functions

Description

This class of special functions constructs model terms with specific properties to be used with asreml syntax.

Usage

dsum(model, levels = NULL, outer = FALSE)

C(obj, contr)

con(obj)obj

lin(obj)

pow(obj, p = 1, offset = 0)

pol(obj, t = 1, init = NA)

leg(obj,t = 1, init = NA)

spl(obj, k = 50, init = NA)

dev(obj, init = NA)

ma(obj)

at(obj, lvls)

and(obj, times = 1)

mbf(obj)

grp(obj)

Arguments

model

A formula of the form ~A+B+...|Z, where A and B define variance matrices for simple or compound model terms, and Z is a simple conditioning factor whose levels identify and determine the number of sub-matrices in the direct sum. The "|" operator is applied associatively and operates with all terms on its left; that is, A+B|C implies (A+B)|C and is equivalent to A|C+B|C.

levels

A list of the length of the number of terms in the left-hand side of model that are separated by "+". The components of levels are vectors of unique values of Z. If there is only one term in the left-hand side of model (or if the context allows, see examples) then levels may be a vector. If levels is numeric, then these are ordinal numbers that refer to the sections defined by Z in unique(Z) order. If NULL then unique(Z) is used (default = NULL).

obj

An object in the data frame.

mbf: A component name from the mbf list argument.
grp: A component name from the group list argument.

contr

An integer vector of contrast coefficients parallel to levels(obj).

p

The exponent in a power function term (pow) (default = 1).

offset

Constant added to obj (default = 0).

t

A value.

pol: The maximum degree of a set of orthogonal polynomials formed from obj. If negative, the intercept in the polynomial is omitted (default = 1).
leg: The maximum degree of a set of Legendre polynomials formed from obj. If negative, the intercept is omitted (default = 1).

k

The number of equally-spaced knot points for a cubic smoothing spline. If zero or omitted, k is set to asreml.options()$knots (default 50).

init

Optional initial value for the default identity variance model idv when used in the random formula.

lvls

Vector of levels of the conditioning factor (obj) that define the conditioning covariates formed by at. If numeric, lvls indexes the levels vector of obj; that is levels(obj)[lvls].

times

It multiplies (may be non-integer) of the design matrix for obj are added to the preceding design matrix (default = 1).

Functions

asr_dsum(): Direct sum structures for residual models.
asr_C(): Defines specific contrasts for a factor.
asr_con(): Sum to zero constraints.
asr_lin(): Create a variate from obj (usually factors).
asr_pow(): Create the model term: (obj+offset)^p.
asr_pol(): Orthogonal polynomials.
asr_leg(): Legendre polynomials.
asr_spl(): Cubic smoothing spline (random component).
asr_dev(): Spline deviations; create a factor from the variate obj.
asr_ma(): Construct a term with a moving-average of order 1 design matrix from obj.
asr_at(): Form a conditioning covariate from obj for each level of obj specified in the lvls argument.
asr_and(): Multiply the design matrix for obj by times and add it to the preceding design matrix.
asr_mbf(): Create a model term from covariates not stored in data.
asr_grp(): Create a model term from covariates held in specific columns of data.

Examples

## Not run: 

## Separable autoregressive residual model at each level of Site
residual = ~dsum(~ar1(Column):ar1(Row) | Site)

## Different residual models at different levels of Site.
## The ordinals in the levels list refer to the unique values of
## Site in unique(Site) order.
residual = ~dsum(~ar1(Column):ar1(Row) + id(Column):ar1(Row) | Site,
                levels = list(c(1, 3), c(2)))

## Equivalent
residual = ~dsum(~ar1(Column):ar1(Row) | Site, levels = c("Site1", "Site3")) +
            dsum(~id(Column):ar1(Row)  | Site, levels = c("Site2"))

## "biological" Date (as factor) within Plot
residual = ~dsum(~ar1(Date) | Plot)

## "explicit" times (Date as variate)
residual = ~dsum(~exp(Date) | Plot)

## End(Not run)

[Package asreml version 4.2.0.480 Index]