冗余分析验证

最新推荐文章于 2024-09-23 08:09:44 发布

1199Ly

最新推荐文章于 2024-09-23 08:09:44 发布

阅读量1k

点赞数

分类专栏：生态学数据分析文章标签： r语言数据分析生物学

本文链接：https://blog.csdn.net/qq_40458594/article/details/104920134

版权

生态学数据分析专栏收录该内容

1 篇文章 0 订阅

订阅专栏

群落排序RDA验证

典范分析结合了排序和回归的思想，其包含一个响应变量矩阵Y和解释变量矩阵X，是与多元回归密切相关的：
在这里插入图片描述
多元回归模型预测 $\hat{Y_i}$ 与Y的排序是不同的，二者相关系数的平方根就是多元回归模型的决定系数。因此，多元回归模型建立了 $\hat{Y_i}$ 排序与Y排序之间的对应，因为回归拟合值排序受限于X，并且与X是最优的且线性相关的。这个特性在典范分析中也很重要。这种限制是在最小均方（least-square）意义上最优的，意味着多元线性回归模型得到了最大的 $R^2$ （决定系数）。因此，典范分析结合了排序和回归的方法，在X的约束下，形成与X密切相关的Y排序。

vagan包进行RDA分析

library(vegan)
library(magrittr)

## Loading required package: permute

## Loading required package: lattice

## This is vegan 2.5-6

spe <- read.csv("2020otuabund.csv",header=T,row.names = 1) %$% .[,-10]
spe <- t(spe)[,1:16]
env <- read.csv("waterc.csv",header=T,row.names = NULL) %$% .[,-(1:2)]

sp <- scale(spe, center = TRUE)
water <- scale(env, center = TRUE)
water <- water[,c(1,2,7,8)]
mod11<-rda(sp ~  TotalC+TotalN+SAL+pH, env)
summary(mod11)

## 
## Call:
## rda(formula = sp ~ TotalC + TotalN + SAL + pH, data = env) 
## 
## Partitioning of variance:
##               Inertia Proportion
## Total          16.000     1.0000
## Constrained     9.832     0.6145
## Unconstrained   6.168     0.3855
## 
## Eigenvalues, and their contribution to the variance 
## 
## Importance of components:
##                         RDA1   RDA2    RDA3    RDA4    PC1    PC2     PC3
## Eigenvalue            6.5226 2.1684 0.78380 0.35745 2.7334 1.7448 1.07726
## Proportion Explained  0.4077 0.1355 0.04899 0.02234 0.1708 0.1091 0.06733
## Cumulative Proportion 0.4077 0.5432 0.59218 0.61452 0.7854 0.8944 0.96173
##                           PC4
## Eigenvalue            0.61228
## Proportion Explained  0.03827
## Cumulative Proportion 1.00000
## 
## Accumulated constrained eigenvalues
## Importance of components:
##                         RDA1   RDA2    RDA3    RDA4
## Eigenvalue            6.5226 2.1684 0.78380 0.35745
## Proportion Explained  0.6634 0.2205 0.07972 0.03635
## Cumulative Proportion 0.6634 0.8839 0.96365 1.00000
## 
## Scaling 2 for species and site scores
## * Species are scaled proportional to eigenvalues
## * Sites are unscaled: weighted dispersion equal on all dimensions
## * General scaling constant of scores:  3.363586 
## 
## 
## Species scores
## 
##            RDA1     RDA2      RDA3     RDA4        PC1      PC2
## OTU72  -0.15864 -0.26192  0.432768 -0.04309 -0.5901195 -0.02104
## OTU401 -0.15190  0.41764 -0.070863  0.11729  0.4254556  0.32065
## OTU822  0.41641 -0.27951  0.074570 -0.18489 -0.0078372  0.55600
## OTU334 -0.67922  0.07835  0.088613  0.04723  0.3248174  0.20553
## OTU815  0.60617 -0.41794  0.192550  0.09392 -0.0280372  0.05853
## OTU828  0.64631 -0.10303  0.240456 -0.02145  0.2288193 -0.29497
## OTU461 -0.68355 -0.32370  0.189123  0.06208 -0.0008486  0.15978
## OTU361 -0.70671 -0.23541 -0.055085 -0.03725  0.0061413 -0.04282
## OTU912  0.24139 -0.18079 -0.250814 -0.27749 -0.5223089  0.40881
## OTU83   0.46548  0.35236  0.115120  0.20873 -0.3058090 -0.44886
## OTU638  0.52096 -0.43189 -0.328720  0.17608  0.3003277 -0.10085
## OTU75  -0.78599 -0.08045  0.023342 -0.03683 -0.0086104  0.02925
## OTU643  0.70966  0.14607  0.147374 -0.07273  0.1339911 -0.36211
## OTU868  0.59002  0.13248  0.001474  0.01602 -0.4618344 -0.33728
## OTU673  0.09015 -0.67743 -0.090516  0.14908  0.3240885 -0.06191
## OTU651  0.39161  0.11345  0.011508 -0.07418  0.6722584 -0.18163
## 
## 
## Site scores (weighted sums of species scores)
## 
##       RDA1      RDA2     RDA3    RDA4      PC1      PC2
## H1 -1.5505 -0.004236 -1.50882 -1.6763 -0.39559 -0.61280
## H2 -1.6591  1.878655 -0.59495  0.7410  0.62023  1.45094
## H3 -1.4895 -2.330115  1.21685  0.9664  0.35000  0.01861
## M1  0.2531 -0.405126 -0.35016 -1.7639 -0.20189  0.60985
## M2  0.7313  1.079749  2.39744 -0.2944 -2.09920 -0.98534
## M3  0.3798  0.908631  0.95080  1.3976 -0.06154 -1.47745
## N1  1.2253 -0.936031 -1.90587  1.5087 -0.37623  0.01875
## N2  1.2686  0.597322 -0.28386  2.5384  2.44508 -1.06512
## N3  0.8410 -0.788848  0.07857 -3.4175 -0.28087  2.04257
## 
## 
## Site constraints (linear combinations of constraining variables)
## 
##       RDA1    RDA2      RDA3    RDA4      PC1      PC2
## H1 -1.5880  0.3692 -1.461026 -1.2108 -0.39559 -0.61280
## H2 -1.3262  1.7883 -0.071104  1.1527  0.62023  1.45094
## H3 -1.4864 -2.2507  1.397221  0.8040  0.35000  0.01861
## M1  0.4777 -0.9880 -0.510043 -1.8351 -0.20189  0.60985
## M2  0.6579  0.8215  1.415913  0.4964 -2.09920 -0.98534
## M3 -0.2408  0.5166  0.006138 -0.3785 -0.06154 -1.47745
## N1  1.2080 -0.8749 -2.035604  1.8738 -0.37623  0.01875
## N2  1.2022  0.4497  0.526211 -0.1378  2.44508 -1.06512
## N3  1.0956  0.1684  0.732294 -0.7648 -0.28087  2.04257
## 
## 
## Biplot scores for constraining variables
## 
##           RDA1     RDA2     RDA3     RDA4 PC1 PC2
## TotalC -0.8196 -0.40984  0.33625  0.21740   0   0
## TotalN -0.3354  0.35041  0.05323  0.87287   0   0
## SAL    -0.7938 -0.38601 -0.44072  0.16332   0   0
## pH      0.9945 -0.07912 -0.02377 -0.06469   0   0

原理验证

#计算拟合值排序矩阵、响应变量排序矩阵
XtX = t(water) %*% water
invers = solve(XtX)
fitY <- water %*% invers %*% t(water) %*% sp
resY <- sp - fitY

U <- eigen(cov(fitY))$vectors
Ures <- eigen(cov(resY))$vectors

FY <- sp %*% U
ZY <- fitY %*% U
NYres <- resY %*% Ures

B = invers %*% t(water) %*% sp
C = B %*% U

var <- eigen(cov(fitY))$values     #各个典范轴方差贡献
var

##  [1]  6.522593e+00  2.168415e+00  7.837967e-01  3.574501e-01  2.949718e-16
##  [6]  1.517841e-16  1.197628e-16  1.173570e-16  6.710281e-17  3.868146e-17
## [11]  9.735283e-18 -1.314287e-17 -2.619791e-17 -6.737024e-17 -1.219536e-16
## [16] -2.560984e-16

varres <- eigen(cov(resY))$values
varres

##  [1]  2.733406e+00  1.744801e+00  1.077257e+00  6.122818e-01  4.778527e-16
##  [6]  1.540008e-17 -9.321812e-18 -1.569721e-17 -1.704919e-17 -3.053562e-17
## [11] -4.060141e-17 -4.277304e-17 -4.325843e-17 -6.131568e-17 -7.034357e-17
## [16] -1.931979e-16