User Tools

Site Tools


en:varpart

This is an old revision of the document!


Constrained ordination

Variation partitioning

Note: variation partitioning is sometimes also called commonality analysis in reference to the common (shared) fraction of variation (Kerlinger & Pedhazur 1973). It is also a synonym to variance partitioning1).

In case we have two or more explanatory variables, one may be interested in variation in species composition explained by each of them. If some of these explanatory variables are correlated, one must expect that variation explained by the first or the other variable cannot be separated - it will be shared.

Invalid Link
Venn's diagram, showing variation explained by two environmental variables (or two sets of environmental variables) and coded names of these fractions. Created by function showvarpart (2) from library vegan.

The way how to approach this problem is variation partitioning, when variation explained by each variable (or set of variables) independently (marginal variation) is partitioned into variation attributable purely to given environmental variable (partial variation), and shared variation attributable to two or more variables.

Variation can be partitioned into individual variables (e.g. variation explained by soil pH vs variation explained by soil Ca) or into groups of variables (e.g. soil variables vs climatic variables).

Result can be visualized using Venn's diagram (see figure on the right). Meaning of the fractions in Venn's diagram is the following2):

  • [a] - partial variation explained by variable 1 (i.e. variation explained by this variable after removing variation explained by the second variable);
  • [c] - partial variation explained by variable 2;
  • [b] - shared variation explained by both variables (cannot be decided to which of them should be attributed, and is a result of correlation between both variables);
  • [a+b] - simple (or marginal) variation explained by variable 1 (i.e. variation this variable would explain if it is as the only explanatory variable in the model);
  • [b+c] - marginal variation explained by variable 2;
  • [d] - unexplained variation.

If the variation is partitioned among groups with the same number of variables (e.g. two soil and two climatic variables), than variation explained by each group is comparable without adjustement. However, if groups contain different numbers of variables, variation explained by not adjusted R2 is not comparable, since R2 tends to increase with the number of explanatory variables. Here, use of adjusted R2 is recommended.

The library vegan offers function varpart, which can partition variation among up to four variables (or groups of variables). Note that varpart is based on redundancy analysis (rda) and uses adjusted R2 to express explained variation. The reason for using only rda is that in R, there is still no function available to calculate adjusted R2 for unimodal ordination methods (like cca).

R functions

  • varpart (library vegan) - variation partitioning (using linear constrained ordinatino - rda) among up to four matrices of environmental variables. First argument (Y) is dependent (usually species composition) matrix (but could be also only one variable - in that case varpart is conductin linear regression). Next arguments (up to four) are (groups of) explanatory variables. Uses either formula interface (with ~) or matrices.

Examples of use

Use of varpart function (using vegetation data from Carpathian wetlands)

Example: how much variation in species data (vasc.hell) explains variables Mg and Ca (from chem)?

# Carpathian wetlands - import data
vasc <- read.delim ('http://www.davidzeleny.net/anadat-r/data-download/vasc_plants.txt', row.names = 1)
chem <- read.delim ('http://www.davidzeleny.net/anadat-r/data-download/chemistry.txt', row.names = 1)
 
# transform data using Hellinger transformation
vasc.hell <- decostand (vasc, 'hell')
 
# upload library vegan if not yet done
library (vegan)
 
# apply function varpart
vp1 <- varpart (vasc.hell, ~ Mg, ~ Ca, data = chem)

In this function, the first is coming the species matrix, than explanatory matrices (or variables) - if using formula interface, each has to start with tilda (~). If these variables are part of matrix of explanatory variables, you need to specify the environmental matrix in argument data =.

Result is:

vp1
Partition of variation in RDA

Call: varpart(Y = vasc.hell, X = ~Mg, ~Ca, data = chem)

Explanatory tables:
X1:  ~Mg
X2:  ~Ca 

No. of explanatory tables: 2 
Total variation (SS): 39.151 
            Variance: 0.56741 
No. of observations: 70 

Partition table:
                     Df R.squared Adj.R.squared Testable
[a+b] = X1            1   0.11987       0.10693     TRUE
[b+c] = X2            1   0.13898       0.12632     TRUE
[a+b+c] = X1+X2       2   0.16357       0.13860     TRUE
Individual fractions                                    
[a] = X1|X2           1                 0.01228     TRUE
[b]                   0                 0.09465    FALSE
[c] = X2|X1           1                 0.03167     TRUE
[d] = Residuals                         0.86140    FALSE
---
Use function 'rda' to test significance of fractions of interest

Alternative way how to use the function is not using formula interface:

varpart (vasc.hell, chem$Mg, chem$Ca)

or, in case you use not-transformed data and you want them to be transformed (using decostand function):

varpart (vasc, chem$Mg, chem$Ca, transfo = 'hell')

(argument transfo is passed into function decostand together with species data).

plot.varpart (library vegan)

I can also plot directly so called Venn's diagram:

plot (vp1)

Note - the plotting function is called plot.varpart, but if I use generic plot function on object vp1 (which is of class “varpart”), I don't have to specify the whole name.


1)
As to the distinction between variance and variation, Legendre & Legendre (2012) note: “The term variation, a less technical and looser term than variance, is used because one is partitioning the total sum of squared deviations of y from its mean (total SS). In variation partitioning, there is no need to divide the total SS of y by its degrees of freedom to obtain the variance s2y.” The first edition of the book Multivariate Analysis of Ecological Data using CANOCO (Lepš & Šmilauer 2003) was using the term variance partitioning, while in the second edition (Šmilauer & Lepš 2014), authors adopted the term variation partitioning, noting: “It was called variance partitioning in the original paper, but we prefer, together with Legendre & Legendre 2012, the more appropriate name referring to variation, as we include also unimodal ordination methods in our considerations.”
2)
Note that in CANOCO 5, the coding of the fraction follows different logic - [c] is shared variation, and [a] or [b] are partial fractions; meaning of [d] remains the same (unexplained variation).
en/varpart.1429441148.txt.gz · Last modified: 2017/10/11 20:36 (external edit)