- New zero-handling procedure
- If
shrink = TRUE, bin frequencies are estimated by James-Stein-type shrinkage
- If
- New objective functions
- Add
asSLRsupport toobjective.keepWADISTby comparing with Euclidean distance objective.maxDSKLmaximizes the RDA of a PCoA of the SKL divergences
- Add
- New objective function
objective.keepSKLpreserves the symmetric Kullback-Leibler (SKL) divergences
- New objective function
objective.keepWADISTpreserves the weighted Aitchison distances
- Update information objectives
objective.keepEntropynow recloses the amalgamation (as it should)objective.maxDEntropynow recloses the amalgamation (as it should)
- New backend information functions
shannonfunction calculates Shannon's indexuniformityfunction calculates a normalized Shannon's index
- New objective functions
objective.keepEntropypreserves the relative entropy of the amalgamsobjective.maxDEntropymaximizes the between-group difference in amalgam entropy
- Update
amalgamfunctionprepareArgsnow forces a re-closure of the data
- Switch
objective.maxRDAandobjective.maxRDA2(again)
- Switch
objective.maxRDAandobjective.maxRDA2
- Update
objective.maxRDAandobjective.maxRDA2- Now sums inertia across all RD axes (as it should)
- Update
objective.maxRDA2method- Like
objective.maxRDAbut makes "z" the community matrix
- Like
- Update
objective.maxRDAmethod- Handle rare case when constrained axes explain no variance
- Add check for "z" argument
- Have
prepareArgscoerce "z" as adata.frame - Update
objective.maxRDAmethod- Now save the intermediate data to debug
- Update
objective.maxRDAmethod- Add
tryCatchto handle the Lapack routine 'dgesdd' error - This error occurs when the SVD fails to converge
- Add
- New
as.slrfunction turns components into a set of summed log-ratios - Update
amalgam- New
asSLRargument passed to objective functions - Objective functions now support SLRs
- Update
plotmethod for SLRs
- New
- Update
amalgam- Zeros now replaced with
zCompositions::cmultRepl - Correctly coerce input as matrix
- Zeros now replaced with
- New scalar product methods
- New
adotcomputes Aitchison scalar product for two vectors - New
amatcomputes Aitchison scalar product for two matrices - New unit tests
- New
- Update
plot.amalgam- Add argument to select which amalgams to show in ternary plot
- Move
zCompositionsto Suggests
- New
intmethodsint.fromNbitconverts a binary vector into integers
- New
weightmethodsweight.Nto1builds a many-to-one amalgamation matrixweight.NtoNbuilds a many-to-many amalgamation matrix
- New
objectivemethodsobjective.keepDistseeks to preserve Aitchison distanceobjective.maxRDAseeks to maximize constraint of RDA
- New
amalgamwrapper- This function runs a genetic algorithm for the objective
print.amalgamshows the object structureplot.amalgammakes 4-panel figure