[Vegan-commits] r408 - branches/1.13/inst

noreply at r-forge.r-project.org noreply at r-forge.r-project.org
Mon Jun 9 10:04:12 CEST 2008


Author: jarioksa
Date: 2008-06-09 10:04:12 +0200 (Mon, 09 Jun 2008)
New Revision: 408

Modified:
   branches/1.13/inst/NEWS
Log:
NEWS update for release 1.13-1

Modified: branches/1.13/inst/NEWS
===================================================================
--- branches/1.13/inst/NEWS	2008-06-09 07:19:54 UTC (rev 407)
+++ branches/1.13/inst/NEWS	2008-06-09 08:04:12 UTC (rev 408)
@@ -5,14 +5,40 @@
 
 		       CHANGES IN VEGAN 1.13-1
 
+GENERAL
+
     - Helene Wagner joined the vegan team.
 
-    - cca, rda, capscale issue a warning when the rank of the
-      constraints + condition is lower than the rank of the depedent
-      (community) data, because vegan is not prepared to handle these
-      cases properly. This is an ancient issue, but surfaced recently.
-      A real fix is expected in next releases.
+FIXES FOR HANDLING RANKS IN CONSTRAINED ORDINATION
 
+    - cca, rda and capscale had only known one kind of rank: the rank
+      of the ordination result (= number of axes). In fact there are
+      two other types of ranks: the rank of constraints after removing
+      conditions, and the rank of constraints + conditions, where
+      conditions refer to terms "partialled out". This hit those that
+      tried to use rda with single response variable instead of
+      community matrix, and also some partial models were handled
+      wrongly. The changes mainly concern cases where the rank of
+      constraints is higher than the rank of the ordination (more
+      constraints than ordination axes).  The changes have visible
+      effects in following support functions:
+      	  
+    - alias: no superfluous aliasing of terms, partial models aliased
+      correctly.
+
+    - anova: always uses rank of the constraints which also fixes
+      anova(..., by = "terms") when single response variable was used.
+
+    - calibrate.cca: identifies cases when the rank of constraints is
+      higher than the rank of ordination and refuses to analyse these.
+
+    - extractAIC: uses rank of constraints for degrees of freedom.
+
+    - predict: predict(..., type = "lc", newdata = somedata) works in
+      partial analysis.
+
+OTHER FIXES
+
     - intersetcor and inertcomp now check that input really is from
       constrained ordination instead of giving obscure error messages.
 
@@ -25,13 +51,12 @@
     - mso: plot.mso replaced with msoplot, and plot function now
       displays the ordination scatter plot. Function msoplot collapses
       distances larger than half of the maximum distance into a single
-      distance class. Default output shows the variogram printout.
+      distance class. Printed result shows the variogram data.
 
     - ordicluster, ordiellipse, ordispider, orglspider, ordisurf,
       factorfit and vectorfit could fail with non-vegan ordination
       objects in R 2.7.0 whose weights.default function gave error if
-      object had no weights (used to return NULL in R 2.6.2 and
-      earlier).
+      object had no weights (work in R 2.6.2 and earlier).
 
     - taxa2dist: has a "method" name for distances.
 



More information about the Vegan-commits mailing list