**1. Highlights**

● 32 new procedures

● density plots of large data sets (DXYDENSITY)

● p-rep (i.e. partially-replicated) designs (AFPREP, CDNPREP)

● random regression forests (BRFOREST. BRFDISPLAY. BRFPREDICT)

● checks of REML residuals (VCHECK)

● standard errors for REML residuals and fitted values (VFRESIDUALS)

● generalized heritability (VHERITABILITY)

● line-by-tester analysis (VALINEBYTESTER, VLINEBYTESTER)

● nearest-neighbour analysis (VNEARESTNEIGHBOUR)

● selection by genetic similarity with a target genotype, and calculation of selection indexes (QBESTGENOTYPES, QSELECTIONINDEX)

● recovery from unsuccessful model fitting in automatic REML analysis (VARECOVER)

● automatic export of ANOVA output to R (A2RDA)

**2. What’s new**

2.1 Directives

None yet

2.2 Procedures

AFMEANS forms tables of means classified by ANOVA treatment factors.

AFPREP searches for an efficient partially-replicated design.

ARETRIEVE retrieves an ANOVA save structure from an external file.

ASTORE stores an ANOVA save structure in an external file.

A2RDA saves results from an analysis of variance in R data frames.

BCKEEP saves information from a classification tree.

BKKEEP saves information from an identification key.

BRFDISPLAY displays information about a random regression forest.

BRFOREST constructs a random regression forest.

BRFPREDICT makes predictions using a random regression forest.

BRKEEP saves information from a regression tree.

CDNPREP constructs a multi-location partially-replicated design using CycDesigN.

DBINFORMATION loads information on the tables and columns in an ODBC database, PC Windows only.

DXYDENSITY produces density plots for large data sets.

FBETWEENGROUPVECTORS forms variates and classifying factors containing within-group summaries to use e.g. in a between-group analysis.

FPARETOSET forms the Pareto optimal set of non-dominated groups.

FREGULAR expands vectors onto a regular two-dimensional grid.

FRTPRODUCTDESIGNMATRIX forms summation, or relationship, matrices for model terms.

QBESTGENOTYPES sorts individuals of a segregating population by their genetic similarity with a defined target genotype, using the identity by descent (IBD) information at QTL positions for one or more traits.

QSELECTIONINDEX calculates (molecular) selection indexes by using phenotypic information and/or molecular scores of multiple traits.

QMVESTIMATE replaces missing molecular marker scores using conditional genotypic probabilities.

TABINSERT inserts the contents of a sub-table into a table.

VALINEBYTESTER provides combinabilities and deviances for a line-by-tester trial analysed by VABLOCKDESIGN or VAROWCOLUMNDESIGN.

VARECOVER recovers when REML, is unable to fit a model, by simplifying the random model.

VASDISPLAY displays further output from an analysis by VASERIES.

VASKEEP copies information from an analysis by VASERIES into GenStat data structures.

VCHECK checks standardized residuals from a REML analysis.

VDFIELDRESIDUALS display residuals from a REML analysis in field layout.

VFRESIDUALS obtains residuals, fitted values and their standard errors from a REML analysis.

VHERITABILITY calculates generalized heritability for a random term in a REML analysis.

VLINEBYTESTER analyses a line-by-tester trial by REML.

VNEARESTNEIGHBOUR analyses a field trial using nearest neighbour analysis.

2.2 Functions

None.

**3. What’s changed**

Most of the changes are compatible with Release 16, the previous release. There are a few commands, however, where new options or parameters have been inserted into the existing lists. These may cause problems in statements where option or parameter names have been omitted or abbreviated (see Section 1.7.1 of Part 1 of the Guide to the GenStat Command Language for details). To avoid any difficulty, the name of the option/parameter after the new option/parameter should be given explicitly, and not abbreviated to fewer than four characters.

Any command, where changes in Release 17 may cause incompatibilities in existing programs, is marked in Sections 3.1 and 3.2 by the symbol †. The full details are given in Section 3.4.

3.1 Directives

DEVICE now allows you to control whether the plotting of each graph must be completed before on the device any other commands are executed.

TXCONSTRUCT can now form texts from formulae.

3.2 Procedures

†AGRAPH now allows you to include additional axis markings, transformed onto another scale, on the right-hand side of the y-axis.

†AMCOMPARISON, †AUMCOMPARISON and †VMCOMPARISON can now save the differences between the pairs of means, and the widths of their confidence intervals.

†AREPMEASURES now allows you to control the orders of the contrasts that are fitted.

†ASPREADSHEET now allows you to specify factors within the tables of means whose levels are assumed to be equal for the two means, when calculating effective standard errors, and can save standardized residuals.

†ASTATUS can now save the weight variate from the most recent ANOVA.

†AUGRAPH now allows you to include additional axis markings, transformed onto another scale, on the right-hand side of the y-axis.

†BCIDENTIFY can now provide identifications for specimens with some of the necessary x-values missing or unavailable, and you can now specify the values to use for the explanatory variables.

BCFIDENTIFY now allows you to specify the values to use for the explanatory variables.

BIPLOT now allows you to save more than 2 dimensions.

†BREGRESSION now allows you to define a stopping rule based on the number of observations at a node.

†BRPREDICT can now provide predictions for units when some of the necessary x-values are missing or unavailable, and you can now specify the values to use for the explanatory variables.

†CDNBLOCKDESIGN and †CDNROWCOLUMNDESIGN now allow you to specify how to weight the importance of the information in the design, and to set different seeds for the design search and randomization.

†CSPRO now creates the data structures locally when called within a procedure, instead of in the main program. You can still create the structures in the main program, by setting option SCOPE=global.

DBARCHART now allows you omit any levels where there are only missing values.

DTABLE now allows you to plot lines between the data values without any points at the data values themselves. Also, you can now include additional axis markings, transformed onto another scale, on the right-hand side of the y-axis.

GGEBIPLOT now allows you to to specify or save the groups used for the megaenvironment setting of SCPLOT.

HGGRAPH – when plotting on the scale of the linear predictor, you can now include axis markings, back-transformed onto the natural scale, on the right-hand side of the y-axis.

HGPREDICT now allows you to provide identifiers for the new factors that are defined when LEVELS are specified.

†IMPORT now creates the data structures locally when called within a procedure, instead of in the main program. You can still create the structures in the main program, by setting option SCOPE=global.

MOVINGAVERAGE can now plot predictions, with the Holt-Winters method, and save their upper and lower confidence limits.

†PROBITANALYSIS can now save the estimates of the lethal doses, and their confidence intervals.

RGRAPH – when plotting on the scale of the linear predictor, you can now include axis markings, back-transformed onto the natural scale, on the right-hand side of the y-axis.

RPHKEEP can save, and RPHDISPLAY and RPHCHANGE can print, the number of degrees of freedom in the model.

SETDEVICE now allows you to control whether the plotting of each graph must be completed on the device before any other commands are executed.

SUBSET has a new option NULL to indicate if the subset contains no units. If this is set, SUBSET suppresses the fault that it normally gives for an empty subset.

TRELLIS now allows you to control whether the same or different pens are used to plot each y-variate.

†VABLOCKDESIGN has a new setting of the RSTRATEGY option that tries to fit the full random model and then, if that fails, tries simpler models until it finds one that can be fitted successfully; this is now the default. You can also control the terms whose effects and means are printed, and the type of standard errors that are presented.

†VAMETA now allows you to request that a simpler random model is fitted, if REML is unable to fit the random model that has been specified. You can also control the terms whose effects and means are printed, and the type of standard errors that are presented.

†VARANDOM now allows you to exclude models with bound or aliased variance parameters. You can also control the terms whose effects and means are printed, and the type of standard errors that are presented.

†VAROWCOLUMNDESIGN has new settings of the RSTRATEGY option that try simpler models if the full random model or a user-specified random model fails; the default is now to try the full model, and then simpler models if that fails. You can also control the terms whose effects and means are printed, and the type of standard errors that are presented.

†VASERIES now has an RSTRATEGY option to allow you to select the strategy to use to determine the random model. You can also control the terms whose effects and means are printed, and the type of standard errors that are presented.

VFMODEL now allows you to avoid overwriting the original model structure when making changes, and to set the experiments factor for a meta analysis.

†VGRAPH can now plot predicted means at the values of a variate as the x-dimension of the plot, and it now uses VPREDICT, instead of VKEEP, to obtain the means and their standard errors. Also, you can now include additional axis markings, transformed onto another scale, on the right-hand side of the y-axis.

†VFSTRUCTURE now allows you to define residual models for experiments in a meta analysis.

†VLSD and †VMCOMPARISON now allow you to specify the number of degrees of freedom to use to calculate the t-statistics and tests.

VPLOT can now plot standardized residuals.

3.3 Functions

None.

3.4 Incompatibilities

AGRAPH procedureoption YTRANSFORM inserted before SAVE.

AMCOMPARISON procedureparameter DIFFERENCES inserted before LABELS.

AREPMEASURES procedureoptions CONTRASTS and DEVIATIONS inserted before FPROBABILITY.

AUMCOMPARISON procedureparameter DIFFERENCES inserted before LABELS.

AUGRAPH procedureoption YTRANSFORM inserted before SAVE.

ASPREADSHEET procedureoptions EQFACTORS and RMETHOD and inserted before SPREADSHEET.

ASTATUS procedureparameter WEIGHTS inserted before SAVE.

BCIDENTIFY procedureoption TERMINALNODES now saves a pointer, instead of a variate, to allow for the fact that several nodes may have been used to form the identification when some x-values are missing or unavailable. Restrictions are now ignored.

BREGRESSION procedureoption NSTOP inserted before OWNBSELECT.

BRPREDICT procedureoption TERMINALNODES now saves a pointer, instead of a variate, to allow for the fact that several nodes may have been used to calculate the prediction when some x-values are missing or unavailable. Restrictions are now ignored.

CDNBLOCKDESIGN procedureoption WEIGHTS inserted before SEED.

CDNROWCOLUMNDESIGN procedureoptions WEIGHTS and RCWEIGHTS inserted before SEED.

CSPRO proceduredefault of SCOPE option changed to local instead of global.

IMPORT proceduredefault of SCOPE option changed to local instead of global.

PROBITANALYSIS procedureoption of CIPROBABIITY inserted before LOGBASE; this replaces the previously, and erroneously, undocumented option %CONFIDENCE (also just before LOGBASE).

VABLOCKDESIGN procedureoptions PTERMS and PSE inserted before MVINCLUDE; new default, allfeasible, for the RSTRATEGY option.

VAMETA procedureoptions PTRY and PRECOVERY inserted before FIXED; parameter EXIT inserted before SAVE.

VARANDOM procedureoptions PTERMS and PSE inserted before MVINCLUDE.

VAROWCOLUMNDESIGN procedureoptions PTERMS and PSE inserted before MVINCLUDE; new default, allfeasible, for the RSTRATEGY option.

VASERIES procedureoptions PTERMS and PSE inserted before MVINCLUDE; option RSTRATEGY inserted before METHOD; parameter EXIT inserted before SAVE.

VFSTRUCTURE procedureoption EXPERIMENT inserted before TERMS.

VGRAPH procedureoption YTRANSFORM inserted before SAVE.

VMCOMPARISON procedureoptions DFMETHOD, DFGIVEN and FMETHOD inserted before SAVE; parameter DIFFERENCES inserted before LABELS.

VLSD procedureoptions DFMETHOD, DFGIVEN and FMETHOD inserted before SAVE.