dbrda compatibility

[jarioksa]

Branch dbRDA-legendre adds dbrda which is a new implementation of distance-based Redundancy analysis (db-RDA). This is a parallel implementation of RDA (function rda), but not all elements of rda are available. The function decomposes dissimilarities instead of raw data and therefore species scores, WA scores of sites and internal rectangular community data Xbar are unavailable. Because of this, not all vegan methods can handle the result. This issue is a checklist of achieving full compatibility.

• add1
• alias
• anova
• as.mlm: OK after 7d064e5
• biplot: only "monoplot" available. Fails if we do not specifically ask for only display = "sites" because returns only one kind of scores. Not a thing that needs fixing within dbrda.
• bstick
• calibrate: OK after 7d064e5
• coef
• deviance
• drop1
• eigenvals
• extractAIC
• fitted: some distances may be NaN (sqrt() of negative squared dissimilarities). Concerns also -- and more often -- residuals. type = "working" returning Gower G is trivial.
• goodness
• inertcomp
• intersetcor: OK after 7d064e5
• model.frame
• model.matrix
• mso: .NotYetImplemented() -- probably doable, but I don't have time for this now, and leave this for (possible) later implementation.
• nobs
• ordiR2step
• ordiresids
• ordistep
• ordixyplot
• permutest
• plot
• points
• predict Update 25/5/16: type="wa" fixed with 4cde862
• print
• residuals: see fitted (this uses the same function).
• RsquareAdj
• scores
• screeplot
• simulate: don't have an idea how to implement without column scores, and therefore disabled.
• spenvcor: OK after 7d064e5
• stressplot
• summary
• text
• tolerance
• vif.cca
• weights
• update man/capscale.Rd
• update man/cca.object.Rd