CRAN 2.4-4 release

This is a minor release that fixes the following issues

• orditkplot passes CRAN tests.
• anova(<cca-object>, by = "axis") ignored partial terms. Function uses now forward testing which is less dangerously biased than previous marginal tests.
• summary and inertcomp for RDA, CCA and frieds failed if constraints had zero rank.
• meandist labels are no longer cropped in plots.
• Canberra distance in vegdist can now handle negative entries in input.

CRAN release 2.4-3

• Registers C and Fortran routines: no longer NOTE in checking the package.
• Fixes weighting issues with cca result object: changes in centring of biplot scores, in as.mlm methods, and in intersetcor and spenvcor.
• Allows correlation scaling in biplot.rda.
• Allows using species scores in envfit and ordisurf.
• Allows using non-standard variable names in Condition in constrained ordination method.
• Fixes issues in ordiR2step and only allows forward selection.
• Allows configuration of ordispider plots for WA/LC scores in constrained ordination.
• Fixes several issues in tolerance.cca

CRAN release 2.4-2

vegan 2.4-2 was released in CRAN on 17/1/17. The release contains more complete NEWS. The major changes are:

• Fixes permutation tests in distance-based constrained ordination (issue #198, PR #199).
• Fixes internal scaling of scores in some constrained ordination methods.
• Fixes weighting in intersetcor
• Fixes issue #219 in betadisper permutation tests.
• Fixes (invisible) return object of ordicluster.
• Fixes tolerance.cca (issue #216).
• Changes scaling of biplot arrows in constrained ordination plots.
• Adds new argument and improves colour coding of small positive values in tabasco.
• Faster nullmodel generation.
• Faster vegdist.
• More robust handling of formula which also fixes issue #200.
• Updated taxonomy of dune meadow species.

CRAN release 2.4-1

vegan_2.4-1 was released to CRAN. This fixes the following github issues:

• issue #187: diversity with NA. Also checks against negative input data.
• issue #189: goodness.cca failed when a component had only one axis. This radiated into fixes in several other functions.
• issue #190: goodness.cca regression of ignoring choices argument.
• issues #192 and #193: more robust handling of factors with missing values. Originally reported as an error for betadisper, but radiated into similar fixes in many other functions. The error report #192 was later rectified by fixing R.
• issue #195: regression in linestack which failed to handle expression. The bug fix part of #195 was fixed here, but the feature request still remains.

Issue #194 discusses other changes.

Most extensive change outside github issues was modernizing Fortran code. This also was the immediate reason for this release: CRAN people asked us to modernize good old Fortran.

More detailed listing of changes can be found in the NEWS file of the release.

CRAN release 2.4-0

A new major release of vegan was released and is available on CRAN. This release is directly lifted from the current master branch and closes the gap between development and release versions. The release will start the new 2.4 series, and the master branch will be 2.5-0 for the next major release.

As this is a major release, there are numerous large changes. Most important user visible changes are listed in the NEWS with the package. Largest families of changes are:

• distance-based RDA and associated analyses were redesigned to directly decompose dissimilarities. With semimetric indices (such as Bray-Curtis) this means that the components of inertia include negative eigenvalues. The release has a new function dbrda for distance-based RDA while it also keeps the old capscale but with redefined inertia. In addition, we have a new implementation of permutational anova as adonis2 along side the classic adonis. Variation partition can also be based on dissimilarities with similar tools. We also include tools for euclidifying dissimilarities in all cases where negative eigenvalues may appear. The changes touch functions capscale, adonis, betadisper and varpart and their support functions.
• null models are expanded. We have new "curveball" algorithm for binary matrices, and many other functions are streamlined. It is also possible to bind several null models together with new smbind functions. With this function, null models can be stratified by groups of species or sampling units, or it is possible to run models in pieces. The models can be expanded with new simulations, or null models can be simulated in parallel.
• several ordi* functions that add graphical items to ordination plots are now more flexible and easier to parametrize with colours, line types etc. This also concerns functions like plotting betadisper results. There is also a new function ordibar that can plot SD, SE or CI crosses instead of corresponding ellipses, and ordiellipse can draw ellipsoid hulls that enclose all points in each group.
• metaMDS strategies changed so that it is easier to avoid local optima and to see which stopping criterion should be changed to reach the solutions. The MDSrotate function can now rotate a multidimensional MDS solution to separate groups optimally in first dimensions.
• designdist knows now beta-diversity notation so that, e.g., Whittaker's index can be calculated as designdist(BCI, "gamma/alpha - 1", alphagamma = TRUE).
• permustats support functions were improved for analysing permutation test statistics from various vegan objects. Now several permutations can be combined with new c() function, and summary can be used to evaluate the P-values also for combined results.
• Adjusted R-squared can now be analysed for db-RDA and cca.

Parallel to this 2.4-0 release, we also released a new version 1.0-1 of vegan3d with similar improvements and bug fixes.

CRAN release 2.3-5

This release was made to fix problems triggered by changes in R-devel: function prcomp gained a new argument rank., and collateral changes made print summary.eigenvals fail in vegan, and we did not pass all tests in CRAN. The new version of handling eigenvals will work both in R-devel and released R versions.

In addition,we fixed a problem in calibrate.cca which would fail if users specified rank. There are also other small fixes which are mostly not visible to users.

CRAN release v2.3-4

Main reason for this release was that texi2dvi (an external system software) version 6.1 introduced incompatible features that prevented building vegan (issue #158). This release removes texi2dvi dependence in a vignette. All other vignettes are also built and can be accessed with standard R tools. Other changes:

• linestack failed in some cases when the labels argument was used.
• index 19 had a typo on betadiver.
• vegandocs was deprecated: we use now standard R tools.

CRAN release 2.3-3

Bug fixes

• Fix issue #152: monoMDS (and hence metaMDS) failed when some dissimilarities were 1e16 or higher.
• Permutations failed with permute control structures in estaccum, ordiareatest, renyiaccum and tsallisaccum.
• Indices that required at least two sampling units in specpool failed when only one sampling unit was supplied.

New Features

• treeheight, treedist and treedive work (or refuse to work) with anomalous trees with reversed branches or negative heights.
• anosim, bioenv and mrpp have more robust handling of alternative input formats.

More in the NEWS file in the package.

CRAN release 2.3-2

• Fix printing details of gridded permutations
• Fix weighting of points in ordicluster
• ordiellipse ignored weights in calculation of standard errors (kind = "se")
• adonis and capscale accept dissimilarities in symmetric square matrix as an alternative to a "dist" object (issue #147)
• mso accepts "dist" objects as an alternative to coordinates of locations
• lines and points of procrustes analysis gained argument truemean to allow adding items to original plots.
• rrarefy returns non-rarefied communities when users ask samples larger than observation (issue #144)

More verbose description in NEWS

CRAN release 2.3-1

Major changes:

• fix values tied with statistics in permutation tests
• fix automatic model building in step, ordistep and ordiR2step with aliased candidate variables in constrained ordinatin
• fix centroids of non-factor classes in cca, rda & capscale
• fix handling very long expression in metaMDS calls
• pass NA dissimilarities to monoMDS
• better checking of inputs in rarecurve
• user-friendly definition of scaling in constrained ordination
• add permustats to anova.cca tests
• add BCI.env data

More verbose description in NEWS.