我查看了在线论坛和各种论文，对我对 RDA 分析结果的解释感到有些困惑。
我在该条件下运行了带有遗传簇的完整模型，并使用 anova.cca() 函数使用方差分析 (PERMANOVA) 的排列提出了一个具有全局测试的重要模型。
signif.full.c <- anova.cca(gno.rda.c)
signif.full.c
#Permutation test for rda under reduced model
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + lat + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#Model 4 221.0 1.0546 0.007 **
# Residual 100 5239.3
然后我查看 RDA 轴，它们都不重要：
signif.axis.c <- anova.cca(gno.rda.c, by="axis")
signif.axis.c
#Permutation test for rda under reduced model
#Forward tests for axes
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + lat + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#RDA1 1 58.0 1.1078 0.123
#RDA2 1 56.3 1.0740 0.307
#RDA3 1 55.3 1.0549 0.302
#RDA4 1 51.4 0.9816 0.686
#Residual 100 5239.3
但是，看看“边距”排列，它着眼于术语的重要性，我得到了经度和深度的显着结果：
signif.margin.c <- anova.cca(gno.rda.c, by="margin")
signif.margin.c
#Permutation test for rda under reduced model
#Marginal effects of terms
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + lat + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#long 1 56.2 1.0717 0.027 *
# lat 1 53.9 1.0285 0.214
#Depth 2 112.8 1.0762 0.007 **
# Residual 100 5239.3
我从模型中删除了纬度，模型和术语一样重要，但 RDA 轴也不重要：
#Permutation test for rda under reduced model
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#Model 3 167.1 1.063 0.005 **
# Residual 101 5293.2
#Permutation test for rda under reduced model
#Marginal effects of terms
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#long 1 55.9 1.0657 0.039 *
# Depth 2 112.4 1.0719 0.015 *
# Residual 101 5293.2
#Permutation test for rda under reduced model
#Forward tests for axes
#Permutation: free
#Number of permutations: 999
#Model: rda(formula = gen.imp ~ long + Depth + Condition(Clusters), data = gno.clusters, scale = T)
#Df Variance F Pr(>F)
#RDA1 1 57.1 1.0900 0.165
#RDA2 1 56.0 1.0681 0.178
#RDA3 1 54.0 1.0308 0.245
#Residual 101 5293.2
这是否意味着我可以忽略模型重要性和术语重要性，因为 RDA 轴不重要？