1. Highlights
● produced in 2010
● 4 new directives, 29 new procedures, 6 new functions
● trellis plots of tables (DTABLE)
● multiple correspondence analysis (MCORANALYSIS)
● stepwise discriminant analysis (SDISCRIMINATE)
● neural networks (NNFIT, NNDISPLAY, NNPREDICT)
● random classification forests (BCFOREST, BCFIDENTIFY and BCFDISPLAY)
● quantile nonlinear regression (RQNONLINEAR)
● parallel anova and regression (AYPARALLEL and RYPARALLEL)
● easier prediction and plotting of predictions from unbalanced analysis of variance (AUPREDICT and AUGRAPH)
● multiple comparisons in unbalanced anova (AUMCOMPARISON)
● plots of REML effects (VDEFFECTS)
● QTL analysis extensions – association mapping (QASSOCIATION, QEIGENANALYSIS), linkage disequilibrium (QLDDECAY), descriptive statistics and diagnostic plots (QMKDIAGNOSTICS), replacement of missing marker scores (QMVREPLACE), percentage variance accounted for by QTLs (QMVAF)
● positive semi-definite approximations non-positive semi-definite matrices (POSSEMIDEFINITE)
● testing of individual fixed and random terms in hierarchical generalized linear models (HGFTEST and HGRTEST)
● stratified random samples (SVSAMPLE)
● measures of association for circular data (CASSOCIATION)
● tests for circular data with common mean direction or identical distributions (CCOMPARE)
● spectra of spatial point patterns (DPSPECTRALPLOT)
2. What’s new
2.1 Directives
BIDENTIFY identifies specimens using a tree.
NNDISPLAY displays output from a multi-layer perceptron neural network fitted by NNFIT.
NNFIT fits a multi-layer perceptron neural network.
NNPREDICT forms predictions from a multi-layer perceptron neural network fitted by NNFIT.
2.2 Procedures
AUGRAPH plots tables of means from AUNBALANCED.
AUMCOMPARISON performs pairwise multiple comparison tests for means from an unbalanced analysis of variance, performed previously by AUNBALANCED.
AUPREDICT forms predictions from an unbalanced design (after AUNBALANCED).
AYPARALLEL does the same analysis of variance for several y-variates, and collates the output.
BCFDISPLAY displays information about a random classification forest.
BCFIDENTIFY identifies specimens using a random classification forest.
BCFOREST constructs a random classification forest.
CABIPLOT plots results from correspondence analysis or multiple correspondence analysis.
CASSOCIATION calculates measures of association for circular data.
CCOMPARE tests whether samples from circular distributions have a common mean direction or have identical distributions.
DPSPECTRALPLOT calculates an estimate of the spectrum of a spatial point pattern.
DTABLE plots tables.
FVSTRING forms a string listing the identifiers of a set of data structures.
HGFTEST calculates likelihood tests for fixed terms in a hierarchical generalized linear model.
HGRTEST calculates likelihood tests for random terms in a hierarchical generalized linear model.
MCORANALYSIS does multiple correspondence analysis.
POSSEMIDEFINITE calculates a positive semi-definite approximation of a non-positive semi-definite symmetric matrix.
QASSOCIATION (now QSASSOCIATION) performs marker-trait association analysis in a genetically diverse population.
QEIGENANALYSIS uses principal components analysis and the Tracy-Widom statistic to find the number of significant principal components to represent a set of variables.
QLDDECAY estimates linkage disequilibrium (LD) decay along a chromosome.
QMKDIAGNOSTICS generates descriptive statistics and diagnostic plots of molecular marker data.
QMVAF calculates percentage variance accounted for by QTL effects in a multi-environment analysis.
QMVREPLACE replaces missing marker scores with one of the scores of the most similar genotypes.
RQNONLINEAR fits and plots quantile regressions for nonlinear models.
RYPARALLEL fits the same regression model to several response variates, and collates the output.
SDISCRIMINATE selects the best set of variates to discriminate between groups.
SVSAMPLE constructs stratified random samples.
T%CONTROL expresses tables as percentages of control cells.
VDEFFECTS plots one- or two-way tables of effects estimated in a REML analysis.
2.3 Functions
COLBIND joins two matrices side by side.
MINSERT inserts a matrix into another matrix.
ROWBIND joins (i.e. stacks) two matrices vertically.
RQOBJECTIVE calculates the objective function from fitting a quantile linear regression.
VPERCENTILES calculates percentiles across a set of variates.
VQUANTILES calculates quantiles across a set of variates.
3. What’s changed
Most of the changes are compatible with Release 12, 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 13 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 3.3.
3.1 Directives
†AXIS, †XAXIS, †YAXIS and †ZAXIS now allow you to use scientific or engineering format for numbers, and to rotate axis labels.
FITCURVE can now fit the emax and generalized emax growth curves.
| emax | y = α + γ / (1 + exp(-β × (log(x) – μ)) ) + ε |
|---|---|
| gemax | y = α + γ / (1 + τ × exp(-β × (log(x) – μ)))1/τ + ε |
These are similar to the logistic and generalized logistic curves, except that their equations involve log(x) instead of x. They are usually used to model decreasing relationships (i.e. the parameter β in the equation is negative), but GenStat will allow increasing relationships with these curves too.
†PEN can now use bitmaps as symbols, and define transparent colours.
In PREDICT and VPREDICT, you can now supply identifiers for the factors that need to be created to index the tables of predictions when LEVELS are specified.
TXFIND can now insist the subtexts to be found must occur within a single line of the target text; it can also save the column and line of the last character of the subtext within the target text.
3.2 Procedures
†AGRAPH can now split the plots across different pages and can draw smooth curves through the points; in addition, the procedure has been simplified following the inclusion of AUGRAPH and DTABLE, which should now be used for plots from unbalanced analysis of variance or plots of general tables.
†AGSQLATTICE can now form lattice square designs.
ANTORDER, ANTTEST and AREPMEASURES now allow the data to be specified in a single variate (with a time factor to identify when each observation was made).
†APOLYNOMIAL can now form the equations of polynomials at various levels of other factors (when their interactions with the polynomial factor have been included in the analysis).
BCIDENTIFY can now form a matrix with the probabilities that the specimens belong to each group.
BOXPLOT now allows the widths and positions of the boxes along the x-axis to be controlled.
†CORANALYSIS has been updated to provide the method and outputs described in Greenacre (2007), Correspondence Analysis in Practice, second edition.
†DBIPLOT now plots arrows on the biplot axes to represent their loadings (or inverse loadings); the loadings show the approximate contribution of each variable in the first two dimensions.
†DCORRELATION can now include axes (and labels) in the plots.
DSCATTER includes a key in the plots when there are groups.
FILEREAD parameter FGROUPS has two new settings form and leave, that should be used in future instead of yes and no to specify explicitly whether each structure should or should not be defined automatically as a factor. This is to remove the confusion arising from the fact that the default for FGROUPS is check, whereas other options and parameters that have no as a setting, use no as their default. However, for compatibility with earlier programs, the settings yes and no will continue to be recognised as synonyms for form and leave.
†FFRAME allows you specify the starting number of the windows, and to save the list of numbers.
GRTHIN allows you to save an indicator indicating whether each point was included or excluded.
†MAANOVA can now save residuals and fitted values.
†RPROPORTIONAL has been rewritten to fit the model by a direct maximization of the likelihood, using NAG algorithm G12BAF; this is much more efficient for large data sets than the alternative method, used in procedure RPHFIT, which fits a generalized linear model to an expanded data set.
R2LINES can save the intercepts.
SAGRAPES option SCALING has two new settings equal and none, that should be used in future instead of yes and no to specify whether or not to have equal scaling for x and y axes on Drift-Unreliability and Discrimination-Disagreement graphs. However, for compatibility with earlier programs, the settings yes and no will continue to be recognised as synonyms for equal and none.
VMCOMPARISON now allows Bonferroni or Sidak adjustments to be made to the critical probability values.
†VGRAPH can produce trellis plots, split over pages and can draw smooth curves through the points.
3.3 Functions
The existing functions are all unchanged.
|3.4 Incompatibilities
AGRAPH procedure |
options COMBINATIONS and ADJUSTMENT deleted; option DFSPLINE inserted before SAVE; parameters MEANS, BAR and BARDESCRIPTION deleted; parameter PAGEGROUPS inserted before NEWXLEVELS. |
|---|---|
AGSQLATTICE procedure |
parameters ROWS and COLUMNS inserted before UNITS. |
APOLYNOMIAL procedure |
parameter GROUPS inserted before COEFFICIENTS. |
AXIS directive |
parameter LROTATION inserted before NSUBTICKS; parameter VREPRESENTATION inserted before SAVE. |
BASSESS directive |
option SELECTION renamed SELECTED (to remove clash the usage in FIT, DESCRIBE etc) |
CORANALYSIS procedure |
now uses the method of Greenacre (2007, Correspondence Analysis in Practice, second edition) by default, instead of the method of Digby & Kempton (1987, Multivariate Analysis of Ecological Communities). |
DBIPLOT procedure |
option MULTIPLIER inserted before WINDOW. |
DCORRELATION procedure |
option SHOW inserted before NCOLOURS. |
DQMQTLSCAN procedure |
options METHOD and DCHROMOSOMES inserted before TITLE, and YAXUPPER inserted before YLABEL. |
DQSQTLSCAN procedure |
options METHOD and DCHROMOSOMES inserted before WINDOW, and YUPPER inserted before SCREEN. |
EXPORT procedure |
option PRINT added before OUTFILE. |
FILEREAD procedure |
parameter FGROUPS has two new settings form and leave, that should be used in future instead of yes and no (although yes and no will continue to be recognised). |
FFRAME procedure |
options STARTWINDOW and TESTGRAPH inserted before NUMBERING; parameter SWINDOW inserted before SYLOWER. |
KERNELDENSITY procedure |
options PLOT and TITLE inserted before WINDOW. |
MAANOVA procedure |
parameters RESIDUALS and FITTEDVALUES inserted before MEANS. |
MOVINGAVERAGE procedure |
option PLOT, with setting movingaverages, replaces the GRAPHICS option . |
PEN directive |
parameters TSYMBOL, TLINE, TFILL and TAREA inserted before SAVE. |
QSBACKSELECT procedure |
parameters ADD2PREDICTORS and DOMINANCEPREDICTORS inserted before CHROMOSOMES, and DOMSELECTED inserted before WALDSTATISTICS. |
QSQTLSCAN procedure |
parameters ADD2PREDICTORS and DOMINANCEPREDICTORS inserted before CHROMOSOMES. |
RQLINEAR procedure |
parameters reordered to match RKEEP: new order is Y, PRQUANTILES, RESIDUALS, FITTEDVALUES, ESTIMATES, SE, VCOVARIANCE, DF, LOWER, UPPER, LOWFITTEDVALUES, UPPFITTEDVALUES, OBJECTIVE, EXIT. |
RPROPORTIONAL procedure |
completely rewritten (but the old procedure is currently still available, as OLDRPROPORTIONAL). |
R2LINES procedure |
parameter INTERCEPTS inserted before LOWER. |
RQSMOOTH procedure |
RESIDUALS parameter moved to come before FITTEDVALUES to match RKEEP etc. |
RSURVIVAL procedure |
option PLOT (with setting survivorfunction) inserted before GRAPHICS to control whether any plots are done. |
SAGRAPES procedure |
option SCALING has two new settings equal and none, that should be used in future instead of yes and no (although yes and no will continue to be recognised). |
VGRAPH procedure |
options LSDLEVEL and DFSPLINE inserted before SAVE; parameters MEANS, BAR and BARDESCRIPTION deleted; parameters TRELLISGROUPS and PAGEGROUPS inserted before NEWXLEVELS. |
XAXIS directive |
parameter LROTATION inserted before NSUBTICKS; parameter VREPRESENTATION inserted before YOMETHOD. |
YAXIS directive |
parameter LROTATION inserted before NSUBTICKS; parameter VREPRESENTATION inserted before XOMETHOD. |
ZAXIS directive |
parameter LROTATION inserted before NSUBTICKS; parameter VREPRESENTATION inserted before XOMETHOD. |