multivariate statistics in ecology and quantitative ...evol.bio.lmu.de › _statgen › multivariate...
TRANSCRIPT
![Page 1: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/1.jpg)
Multivariate Statistics in Ecology andQuantitative Genetics
Correspondence analysis
Dirk Metzler & Martin Hutzenthaler
http://evol.bio.lmu.de/_statgen
Summer semester 2011
![Page 2: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/2.jpg)
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 3: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/3.jpg)
Correspondence analysis Motivation
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 4: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/4.jpg)
Correspondence analysis Motivation
The dependence of species on environmental variablesis often not linear
(often not even increasing/decreasing)
Example:The reproduction rate of bacteria depends on temperature.Low and high temperatures are not good or even lethal.
![Page 5: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/5.jpg)
Correspondence analysis Motivation
The dependence of species on environmental variablesis often not linear
(often not even increasing/decreasing)
Example:The reproduction rate of bacteria depends on temperature.Low and high temperatures are not good or even lethal.
![Page 6: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/6.jpg)
Correspondence analysis Motivation
The following figure shows an artificial example of abundanciesof some species along the environmental variable ’temperature’.
10 15 20 25 30
020
4060
8010
0
Temperature
Abu
ndan
cy
● ●
● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
● ●●
●●
In this artificial example, the niche around 20◦ is prefered.
![Page 7: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/7.jpg)
Correspondence analysis Motivation
The following figure shows an artificial example of abundanciesof some species along the environmental variable ’temperature’.
10 15 20 25 30
020
4060
8010
0
Temperature
Abu
ndan
cy
● ●
● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
● ●●
●●
In this artificial example, the niche around 20◦ is prefered.
![Page 8: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/8.jpg)
Correspondence analysis Motivation
A good and simple model for niches is the so-calledGaussian response model
Zi = C exp(−(Xi − µ)2
2t2
)where C, µ, t are parameters.
This model issimple: only 3 parametersgood: the Gaussian response model is chosen (among allunimodal distributions) because of the normalapproximation (central limit theorem) which loosly speakingsays that if many independent (or at most weaklydependent) effects contribute to a quantity and no effect isdominating, then the quantity is distributed normally.
![Page 9: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/9.jpg)
Correspondence analysis Motivation
A good and simple model for niches is the so-calledGaussian response model
Zi = C exp(−(Xi − µ)2
2t2
)where C, µ, t are parameters.
This model issimple: only 3 parametersgood: the Gaussian response model is chosen (among allunimodal distributions) because of the normalapproximation (central limit theorem) which loosly speakingsays that if many independent (or at most weaklydependent) effects contribute to a quantity and no effect isdominating, then the quantity is distributed normally.
![Page 10: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/10.jpg)
Correspondence analysis Motivation
Fitted Gaussian response model
10 15 20 25 30
020
4060
8010
0
Temperature
Abu
ndan
cy
● ●
● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
● ●●
●●
Optimum µ = 20, maximum C = 100, tolerance t = 2.
![Page 11: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/11.jpg)
Correspondence analysis Setting
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 12: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/12.jpg)
Correspondence analysis Setting
Correspondence analysis
Given: Data frame/matrix YY [i , ·] are the observations of site i
Y [·, j ] are the observations at species j
Goal: Find associations of species and sitesAssumption: There is a niche dependence of the species on the
environmental variables
The setting is formulated here in terms of species and sites.If you have measured quantities (variables) of some objects,then replace ’site’ by ’object’ and ’species’ by ’variable’.
![Page 13: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/13.jpg)
Correspondence analysis Setting
Correspondence analysis
Given: Data frame/matrix YY [i , ·] are the observations of site i
Y [·, j ] are the observations at species j
Goal: Find associations of species and sites
Assumption: There is a niche dependence of the species on theenvironmental variables
The setting is formulated here in terms of species and sites.If you have measured quantities (variables) of some objects,then replace ’site’ by ’object’ and ’species’ by ’variable’.
![Page 14: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/14.jpg)
Correspondence analysis Setting
Correspondence analysis
Given: Data frame/matrix YY [i , ·] are the observations of site i
Y [·, j ] are the observations at species j
Goal: Find associations of species and sitesAssumption: There is a niche dependence of the species on the
environmental variables
The setting is formulated here in terms of species and sites.If you have measured quantities (variables) of some objects,then replace ’site’ by ’object’ and ’species’ by ’variable’.
![Page 15: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/15.jpg)
Correspondence analysis Setting
Correspondence analysis
Given: Data frame/matrix YY [i , ·] are the observations of site i
Y [·, j ] are the observations at species j
Goal: Find associations of species and sitesAssumption: There is a niche dependence of the species on the
environmental variables
The setting is formulated here in terms of species and sites.If you have measured quantities (variables) of some objects,then replace ’site’ by ’object’ and ’species’ by ’variable’.
![Page 16: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/16.jpg)
Correspondence analysis Mathematical background
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 17: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/17.jpg)
Correspondence analysis Mathematical background
Let Y have M rows and N columns.The Gaussian regression from the motivation subsection wouldlead to finding species scores (u1,u2, · · · ,uN) and site scores(l1, l2, · · · , lM) such that
Y [i , k ] ≈ Ck exp(− (li − uk)
2t2k
)In fact, correspondence analysis does not use this approach buta weighted PCA approach which results in a similarrepresentation.
![Page 18: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/18.jpg)
Correspondence analysis Mathematical background
The approach of correspondence analysis is based on theChi-square statistic which is used for testing the nullhypothesisthat the species do not depend on the sites. In that case wewould have
Y [i , k ]n
≈ Y [i ,+]
n· Y [+, k ]
nwhere
Y [i ,+] =∑
k Y [i , k ] is the row sum,Y [+, k ] =
∑i Y [i , k ] is the column sum and
n = Y [+,+] =∑
i
∑k Y [i , k ] is the total sum.
The Chi-square test statistic is given by
X 2 =∑∑ (Observed − Expected)2
Expected
=∑
i
∑k
(Y [i , k ]/n − Y [i ,+]Y [+, k ]/n2
)2
Y [i ,+]Y [+, k ]/n2
![Page 19: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/19.jpg)
Correspondence analysis Mathematical background
The approach of correspondence analysis is based on theChi-square statistic which is used for testing the nullhypothesisthat the species do not depend on the sites. In that case wewould have
Y [i , k ]n
≈ Y [i ,+]
n· Y [+, k ]
nwhere
Y [i ,+] =∑
k Y [i , k ] is the row sum,Y [+, k ] =
∑i Y [i , k ] is the column sum and
n = Y [+,+] =∑
i
∑k Y [i , k ] is the total sum.
The Chi-square test statistic is given by
X 2 =∑∑ (Observed − Expected)2
Expected
=∑
i
∑k
(Y [i , k ]/n − Y [i ,+]Y [+, k ]/n2
)2
Y [i ,+]Y [+, k ]/n2
![Page 20: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/20.jpg)
Correspondence analysis Mathematical background
Instead of frequencies we now consider probabilities
p[i , k ] := Y [i , k ]/n
and define a matrix Q with entries
Q[i , k ] :=p[i , k ]− p[i ,+] · p[+, k ]√
p[i ,+]p[+, k ]
Now all further steps are just as in PCA with thecentred/normalized matrix Y replaced by the association matrixQ. Again we get a distance biplot and a correlation biplot.
Correspondence analysis assessesthe association between species and sites
(or objects and variables)
![Page 21: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/21.jpg)
Correspondence analysis Mathematical background
Instead of frequencies we now consider probabilities
p[i , k ] := Y [i , k ]/n
and define a matrix Q with entries
Q[i , k ] :=p[i , k ]− p[i ,+] · p[+, k ]√
p[i ,+]p[+, k ]
Now all further steps are just as in PCA with thecentred/normalized matrix Y replaced by the association matrixQ. Again we get a distance biplot and a correlation biplot.
Correspondence analysis assessesthe association between species and sites
(or objects and variables)
![Page 22: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/22.jpg)
Correspondence analysis Example: Mexican plant data
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 23: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/23.jpg)
Correspondence analysis Example: Mexican plant data
Is there a difference in species community amoung fourgroups of pastures?Data of vegetation on pastures in Mexico from the dryseason of 1992.see also
Zuur, Ieno, Smith (2007) Analysing Ecological Data.Springer
![Page 24: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/24.jpg)
Correspondence analysis Example: Mexican plant data
Meaning of the Variables
Variable names with small letters are shortcuts for plant families.ALTITUDE Altitude above sea level in metersFIELDSLOPE Field slope in degrees
AGE Time since forest clearing; index from 1-8representing ages from 6 to 40 years
CATTLEINTENSITY number of cattle per hectarPLAGUE Nominal: 0=no plague, 1=plague (herbivore insects)
in the year before samplingMAXVEGHEIGHT in centimeter
BLOCK Nominal: 1=grama pastures in Balzapote, 2=starpastures in Balzapote, 3=grama pastures in LaPalma, 2=star pastures in La Palma
(many more variables in original data set)
![Page 25: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/25.jpg)
Correspondence analysis Example: Mexican plant data
mplants<-read.table("MexicanPlants.txt",h=T,sep="\t")
species<-mplants[,c("ac","as","co","com","cy","eu",
"grcyn","grresto","la","le","ma","ru","so","vi","vit")]
library(vegan)
mplants_CA<-cca(species)
plot(mplants_CA, scaling=1, cex=2, main="Scaling=1")
round(mplants_CA$CA$tot.chi, 2)
# 0.33
round(mplants_CA$CA$eig[1:2], 2)
# CA1 CA2
# 0.13 0.06
cat("The first two axis explain",
round(sum(mplants_CA$CA$eig[1:2])
/mplants_CA$CA$tot.chi*100), "%", "\n")
# The first two axis explain 57 %
![Page 26: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/26.jpg)
Correspondence analysis Example: Mexican plant data
−2 −1 0 1 2 3
−2
−1
01
Scaling = 1
CA1
CA
2
ac
as
co
com
cy
eu
grcyn
grresto
la
le
ma
ru
so
vi
vit
14
1920
21
22
1315
16
171836
7
11
12
4
5
8
9
10
![Page 27: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/27.jpg)
Correspondence analysis Example: Mexican plant data
> plot(mplants_CA, scaling = 2, main = "Scaling = 2")
![Page 28: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/28.jpg)
Correspondence analysis Example: Mexican plant data
−2 −1 0 1 2 3
−3
−2
−1
01
Scaling = 2
CA1
CA
2
ac
as
co
com
cy
eu
grcyn
grresto
la
lema
ru
so
vi
vit
14
1920
21
22
1315
16
1718
3
6
7
11
12
4
5
8
9
10
![Page 29: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/29.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 30: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/30.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
The position of a species represents the optimum value in termsof the Gausian response model (niche) along the first andsecond axes. For this reason, species scores are representedas labels or points.
Site conditional biplot (scaling=1)The sites are the centroids of the species, that is, sites areplotted close to the species which occur at those sites.Distances between sites are two-dimensionalapproximations of their Chi-square distances. So sitesclose to each other are similar in terms of the Chi-squaredistance.
![Page 31: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/31.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
The position of a species represents the optimum value in termsof the Gausian response model (niche) along the first andsecond axes. For this reason, species scores are representedas labels or points.
Site conditional biplot (scaling=1)The sites are the centroids of the species, that is, sites areplotted close to the species which occur at those sites.Distances between sites are two-dimensionalapproximations of their Chi-square distances. So sitesclose to each other are similar in terms of the Chi-squaredistance.
![Page 32: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/32.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
Species conditional biplot (scaling=2)The species are the centroids of the sites, that is, speciesare plotted close to the sites where they occur.Distances between species are two-dimensionalapproximations of their Chi-square distances. So speciesclose to each other are similar in terms of the Chi-squaredistance.
![Page 33: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/33.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
There is also a joint plot of species and site scores (scaling=3).In this plot distances between sites and distances betweenspecies can be interpreted as the approximations of therespective Chi-square distances. However the relative positionsof sites and frequencies cannot be interpreted. So this biplot isto be used with care if used at all.
![Page 34: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/34.jpg)
Correspondence analysis Site conditional biplot and species conditional biplot
Note:The total inertia (or total variance) in correspondenceanalysis is defined as the Chi-square statistic of thesite-by-species table divided by the total number ofobservations.Points further away from the origin in a biplot are the mostinteresting as these points make a relatively highcontribution to the Chi-square statistic. So the further awayfrom the origin a site is plotted, the more different it is fromthe average site.
![Page 35: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/35.jpg)
Correspondence analysis Example: An artificial example
Contents
1 Correspondence analysisMotivationSettingMathematical backgroundExample: Mexican plant dataSite conditional biplot and species conditional biplotExample: An artificial example
![Page 36: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/36.jpg)
Correspondence analysis Example: An artificial example
Let us look at the following artificial example which is simpleenough so that we can infer site-species correspondence ’byeye’.
Y <- matrix(c(1,0,0,1,1,0,1,1,1),nrow=3); Y
# [,1] [,2] [,3]
# [1,] 1 1 1
# [2,] 0 1 1
# [3,] 0 0 1
myca <- cca(Y)
plot(myca,scaling=1)
plot(myca,scaling=2)
p <- Y/sum(Y)
pr <- apply(p,1,sum)
pc <- apply(p,2,sum)
expec <- as.matrix(pr) %*% t( as.matrix(pc) )
sum( (p-expec)^2/expec ) # = myca$tot.chi = 0.3611
sum(myca$CA$eig/myca$tot.chi) # 1
![Page 37: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/37.jpg)
Correspondence analysis Example: An artificial example
−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0
−1.
0−
0.5
0.0
0.5
1.0
Scaling=1
CA1
CA
2spe1
spe2
spe3
sit1
sit2
sit3
![Page 38: Multivariate Statistics in Ecology and Quantitative ...evol.bio.lmu.de › _statgen › Multivariate › 11SS › ca.pdf · Correspondence analysis Site conditional biplot and species](https://reader033.vdocuments.net/reader033/viewer/2022060210/5f048f827e708231d40e9380/html5/thumbnails/38.jpg)
Correspondence analysis Example: An artificial example
−1.0 −0.5 0.0 0.5 1.0 1.5 2.0
−1.
0−
0.5
0.0
0.5
1.0
Scaling=2
CA1
CA
2
spe1
spe2
spe3
sit1
sit2
sit3