Prepares pedigree information to generate an inverse relationship matrix for use when fitting animal or plant breeding models by
REML (S.A. Gezan & R.W. Payne).
||Whether to match factor values by their levels or their labels (
|SEX = string token||Possible sex categories of parents (
||Value to be treated as unknown in the pedigree factors|
|INVMETHOD = string token||How to represent the
||Individuals on which data have been measured|
||Male parents (or sires) of the progeny|
||Female parents (of dams) of the progeny|
||New individuals factor, with levels standardized for use in
||New males factor, with levels standardized to match those in the
||New females factor, with levels standardized to match those in the
||Pointer containing additional factors, that may be used in the
||Pointer containing new additional factors, with standardized levels|
||Inverse relationship matrix in sparse matrix form|
||Full list of identifiers generated from the individuals and parents|
In the analysis of animal and plant breeding experiments it may be interesting to take account of the parentage of the animals or genotypes. This pedigree information is specified by three factors, one that identifies the individuals for which data are available, and two others that indicate their male parents and their female parents (if available). This information can be used to generate a sparse inverse relationship matrix that can be used by
VSTRUCTURE to define a correlation model of the individual (or animal) effects for use in a
REML analysis. The matrix is formed using the
VPEDIGREE directive. First, however,
VPEDIGREE needs to standardize the factors so that the levels and labels of the individual male and female factors match, and that the levels are in ascending order, with the parents defined in the individuals factor before their offspring. Otherwise
VPEDIGREE will fail. If you are confident that your factors are already standardized, you can call
VPEDIGREE direct (and use VFPEDIGREE instead if that fails).
The factors defining the individuals, the male parents (or sires) and, optionally, the female parents (or dams) in the pedigree data set.are specified by the
FEMALEPARENTS parameters respectively. The
OTHERFACTORS parameter can specify a pointer containing additional factors, involving the individuals in the pedigree, that may also be needed in the
REML models. You can use the
NEWFEMALEPARENTS and the
NEWOTHERFACTORS parameter parameters to save the new standardized factors. Otherwise, the original factors are redefined.
FREPRESENTATION option indicates whether the factor values are to be matched by their levels (the default) or their labels. If the
FEMALEPARENTS factors are being matched by levels, and the number corresponding to each level needs to be redefined, the factors will be given labels to help identify the original values. If
INDIVIDUALS has labels, these will be used. Otherwise the labels will be textual forms of the original levels.
POPULATION option can save the levels of the standardized factors when
FREPRESENTATION=levels, or their labels when
By default, it is assumed that an individual can act as either a male or female parent but not both. Option
SEX=either can be used to specify that individuals can act as both male and female parents. This may be useful, for example, in plant breeding analyses.
Missing values in any of the factors will be treated as coding for unknown individuals. Option
UNKNOWN allows you to specify an additional code to represent unknown individuals. This should be a scalar (e.g. 0 or -1) when
FREPRESENTATION=levels, or a single-valued text (e.g.
The inverse relationship matrix can be saved by the
INVERSE parameter. By default, this is held in a special sparse matrix form (that is, only non-zero values are stored), using a pointer. This is usable in the
VSTRUCTURE directive but not elsewhere in Genstat. The second element of the pointer is a variate storing the non-zero values of the inverse matrix in lower-triangular order. The first element of the pointer is an integer index vector. Alternatively, you can set option
INVMETHOD=full to store the full matrix as a symmetric matrix (which can also be used by
VSTRUCTURE). However, this is not recommended for large pedigrees.
VFPEDIGREE ignores any restrictions on the factors.
Commands for: REML analysis of linear mixed models.
CAPTION 'VFPEDIGREE example'; STYLE=meta " Basic example of animal model" " Read the data: pedigree and response (in the same file)" " with levels all Individuals" FACTOR [NVALUES=5; LEVELS=!(3,4,5,6,7)] Indiv & [LEVELS=!(0,1,3,5)] Sire & [LEVELS=!(0,2,4,6)] Dam VARIATE [NVALUES=5] Size READ Indiv,Sire,Dam,Size; 3 1 0 12.8 4 1 2 14.5 5 3 4 11.2 6 1 4 12.6 7 5 6 9.9 : VFPEDIGREE [FREPRESENTATION=levels; UNKNOWN=0] Indiv; NEWIndivIDUALS=ID_Indiv;\ MALEPARENTS=Sire; NEWMALEPARENTS=ID_Sire;\ FEMALEPARENTS=Dam; NEWFEMALEPARENTS=ID_Dam;\ OTHERFACTORS=!p(Indiv); NEWOTHERFACTORS=!p(f_Indiv) VPEDIGREE [SEX=fixed] ID_Indiv; MALEPARENTS=ID_Sire; INVERSE=AINV;\ FEMALEPARENTS=ID_Dam VCOMPONENTS RANDOM=f_Indiv; CONSTRAINTS=pos VSTRUCTURE [TERMS=f_Indiv] MODEL=fixed; INVERSE=AINV REML [PRINT=model,components,means; PARAM=sigmas] Size