[Vegan-commits] r433 - branches/1.13/inst

[noreply at r-forge.r-project.org]

Author: jarioksa
Date: 2008-06-17 09:05:13 +0200 (Tue, 17 Jun 2008)
New Revision: 433

Modified:
   branches/1.13/inst/NEWS
Log:
NEWS about fixes in would-be reelease 1.13-2

Modified: branches/1.13/inst/NEWS
===================================================================
--- branches/1.13/inst/NEWS	2008-06-17 06:54:09 UTC (rev 432)
+++ branches/1.13/inst/NEWS	2008-06-17 07:05:13 UTC (rev 433)
@@ -5,10 +5,29 @@
 
 		       CHANGES IN VEGAN 1.13-2
 
+    - anova.cca (and permutest.cca) now calculate the residual df as
+      (number of rows) - (rank of constraints and conditions) - 1
+      instead of using the rank of the residual ordination.  With
+      single response variable in 'rda' the degrees of freedom and
+      F-values are now identical to those in 'lm' (linear
+      model). However, the change does not influence significances
+      (P-values), because these have been always been found by
+      permutation, and this change does not influence the order
+      statistic used in permutations.
+
+    - rda (and capscale) use internal scaling constant so that the
+      returned site and species scores with scalings 1, 2 or 3
+      together provide a biplot approximation of the original
+      data. This scaling constant is calculated internally, but now
+      its numerical value is returned as an attribute of 'scores.rda',
+      and 'summary' displays its value. Vignette on "Design decisions
+      and implementation" in vegan explains the calculation of the
+      internal scaling constant.
+
     - summary of capscale identical to the summary of rda.
 
     - scores of rda and capscale with scaling = 0 really return the
-      unmodified scores.	
+      unmodified scores without the scaling constant.	
 
 		       CHANGES IN VEGAN 1.13-1