G2VEXPORT procedure

Forms a dbase file to transfer REML output to Agronomix Generation II (R.W. Payne).

Options

PRINT = strings	Controls printed output (columns); default * i.e. none
REPLICATETERMS = formula	Specifies the term or terms that define the replication in the design
MODEL = formula	Indicates which model terms (fixed and/or random) are to be used in forming the predictions; default * includes all the fixed terms and relevant random terms
OMITTERMS = formula	Specifies terms to be excluded from the MODEL; default * i.e. none
FACTORIAL = scalar	Limit on the number of factors or variates in each term in the models specified by MODEL or OMITTERMS; default 3
PRESENT = identifiers	Lists factors for which averages should be taken across combinations that are present
WEIGHTS = tables	One-way tables of weights classified by factors in the model; default *
ALPHALEVEL = scalar	Alpha value to use when calculating least significant differences; default 0.05
TAIL = scalar	Number of tails in the calculation of least significant differences (1, 2); default 1
SAVE = REML save structure	Save structure for the analysis from which the means &c are to be saved; default * takes the information from the most recent REML analysis

Parameters

MEANTERM = formula	Defines the treatment term whose means are to be saved; no default (must be specified)
OUTFILE = text	Name of the output file (dbf) to form; default * i.e. file not formed

Description

G2VEXPORT can be used after a Genstat REML analysis, to write a dbase file with a table of means and associated information to be loaded into the Agronomix Generation II system (see agronomix.com).

Printed output is controlled by the PRINT option, with settings:

columns	to print the columns of information to be saved.

By default the means and other information are taken from the analysis of the last y-variate to have been analysed by REML. Alternatively, you can take the information from an analysis of another y-variate, by saving a save structure using the SAVE parameter of REML when it is analysed, and then supplying this to G2VEXPORT using its SAVE option.

The MEANTERM parameter specifies a formula defining the term whose means are to be saved; note that only one table of means can be saved in each call of the G2VEXPORT. The OUTFILE parameter specifies the file (assumed to be a dbase file) where the information is to be stored. The means are calculated using the VPREDICT directive. Options MODEL, OMITTERMS, FACTORIAL, PRESENT and WEIGHTS (which all operate exactly as in VPREDICT) are provided to control how this is done.

The ALPHALEVEL option specifies the alpha value to use in the calculation of least significant differences that accompany the table of means (default 0.05), and the TAIL option specifies whether this is to be for a 1 or 2-sided test (default 1).

The REPLICATETERMS option can supply a model formula to specify one or more model terms defining complete replications of the treatments: for example, blocks in a complete randomized block design, or rows and columns in a Latin square.

Options: PRINT, REPLICATETERMS, MODEL, OMITTERMS, FACTORIAL, PRESENT, WEIGHTS, ALPHALEVEL, TAIL, SAVE.

Parameters: MEANTERM, OUTFILE.

Method

The information is mainly obtained using VKEEP and VPREDICT. The first column (called NAME) describes the contents of each row. Then there is a column for every factor in the table of means, indexing the column of means (called AVG) which comes next. The ranks of the means are in the subsequent column (called RANK), and the next column (called CV) saves the standard deviation of the observations on each combination of the levels of the mean factors, expressed as a percentage of their mean. Finally, if the means are unequally replicated there is a column saving the replication of each mean.

At the top of the columns, there is a row for each mean in the table. Then there are some extra rows with the following names (in the NAME column) and information (in the AVG column):

GRAND MEAN	the grand (i.e. overall) mean;
CV	the coefficient of variation for the lowest stratum in which the maximal model term in the table of means (e.g. A.B for an A-by-B table of means) is estimated;
LSD	saves the least significant difference for the table of means if this is the same for all comparisons of means within the table, otherwise this is replaced by three rows with the minimum, average and maximum LSD (Min LSD, LSD and Max LSD);
Residual	the residual mean square for the lowest stratum in which the maximal model term in the table of means is estimated;
SED	saves the standard error of differences for the table of means if this is the same for all comparisons of means within the table, otherwise this is replaced by three rows with the minimum, average and maximum SED (Min SED, SED and Max SED);
Alpha level	alpha level used in the calculation of the LSDs (ALPHALEVEL option);
R Square	the value of R-square for analysis down to the lowest stratum in which the maximal model term in the table of means is estimated (this ensures that any lower strata that represent within-cell replication are ignored);
No. of Reps	saves replication of the table of means if this is the same for every mean in the table, otherwise this is replaced by three rows with the minimum, average and maximum replication (Min no. of Reps, No. of Reps and Max no. of Reps);
RE-RCBD	this rows is include for compatibility with the output that G2AEXPORT constructs following ANOVA, but it is not useful for the unbalanced designs that REML usually analyses (with G2AEXPORT, it is the efficiency factor of the maximal model term in the table of means, e.g. A.B for an A-by-B table of means, expressed as a percentage);
Rep-Msqr	the mean square of the REPLICATIONTERMS;
Heritability	the efficiency factor of the maximal model term in the table of means;
Prob. Entry	the probability of the maximal model term in the table of means, calculated from the Wald statistic if MEANTERM is a fixed term, or from differences of deviances if it is a random term (note: the probability will then test for the inclusion not only of MEANTERM but also for any of its interactions);
Error d.f.	the residual degrees of freedom;
Tail	Number of tails in the calculation of the least significant differences (TAIL option).

Action with RESTRICT

If the Y variate in the REML was restricted, only the units not excluded by the restriction will have been analysed.

See also

Directive: REML.

Procedures: G2AEXPORT, G2AFACTORS.

Commands for: REML analysis of linear mixed models.

Example

CAPTION     'G2VEXPORT example',\
            'Slate Hall Farm data (Guide to REML in Genstat, Section 1.8).';\
            STYLE=meta,plain
SPLOAD      '%gendir%/data/slatehall.gsh'
VCOMPONENTS [FIXED=variety] replicates/(rows*columns)
REML        [PRINT=model,components,means,waldTests; PSE=differences] yield
G2VEXPORT   [PRINT=columns; REP=replicates] variety