问题
I am a newbie to multivariate statistics, so please forgive me if this question is naïve, or if I have missed something important.
I'd like to know how to deal with an unbalanced design when using adonis. I have a data set of microbial communities present in stream sediment and the factors stream (4 levels), US_DS (2 levels) and season (2 levels) (I have other factors as well but sticking to these for the moment). e.g. code below:
dist.L2<-vegdist(L2, mehod=”bray”)
adonis(dist.L2~Season*Stream*US_DS, data=factors, permutations=199)
Unfortunately the design is unbalanced (some sites only have 1 replicate, while others have up to 3).
I have tried running adonis, however because anodis uses type 1 SS the order that I put the factors in makes a large difference, and there is no reason to put one factor before the others. I know that other stats packages get around this by using type II or III sums of squares, and that there are advantages and disadvantages no matter what type of SS you use. The nature of my data means that other more simple stats tests are not really appropriate.
I would like to know how to either use type III errors with adonis, or whether there is a procedure to go through using type I errors and adonis as it is to find an appropriate test. Thanks
回答1:
You may need to do Adonis twice by changing the order of your two factors.
edit the previously supplied reference link was removed because not active any more
来源：https://stackoverflow.com/questions/26768779/vegan-adonis-unbalanced-design-ss-type-ii-or-iii