Hello everyone,
Could someone help me and explain what is the best way to post hoc at PERMANOVA.
Thanks,
Lucas
I´m studying differences in species abundances among 3 points of an estuary (fixed factor), 4 seasons (fixed) and 2 saltmarshes stratum (fixed). PERMANOVA analysis detect some significant differences among factors. I would like to do a post-hoc analysis to show where are de differences within significant factors.
The p.adjust() function works with any vector of p values. You can use a general adjustment such as Holm or other options depending on circumstances. With multiple comparisons the Hochberg correction is pretty powerful and the assumptions (positively correlated test statistics) should hold).
As said elsewhere there is a lot of methods. To see if amongst you 24 or means/values you have significantly different clusters I would advocate the Tukey method, a bit conservative but quite robust in your open setting ( others available I admit).
Have a look at https://www.r-bloggers.com/anova-and-tukeys-test-on-r/ for some methodological hints.
Hi Lucas,
Maybe it is too late to answer your question. On my side, I also struggled with how to perform a post-hoc test in R with non parametric test (adonis). The problem is that with classic post hoc test, you do not consider the total variance of your dataset, but only the variance associated with the pair you test. Consequently, I gave up with adonis and I now use primer6 which is more convenient to perform the post-hoc test while considering the total variance.
Hope it helps.
Hello everyone,
I found the following function that calculates the post hoc with bonferroni correction.
pairwise.adonis
Please use this version. I corrected some small issues.
https://github.com/pmartinezarbizu/pairwiseAdonis
Pedro
Dear Pedro,
Thank you very much! It is a very nice function
hey Can you do a multiple comparison post hoc, with contrasts on 3 categories?
Hi Pedro Martinez Arbizu
Following the recent update adding pairwise comparisons for interactions (https://github.com/pmartinezarbizu/pairwiseAdonis), I can't seem to understand how the pairwise.adonis2 function is providing this (i am not interested in main effect pairwise comparisons)?
pairwise.adonis2(Tvar_dw^0.25 ~ DEPTH_BIN * ISLAND , strata = Null)
Tvar_dw^0.25 = dataframe
DEPTH_BIN = factor with 3 levels
ISLAND = factor with 5 levels
Using adonis2, i have a significant interaction. Therefore, I am expecting to receive an output of pairwise comparisons of the levels of my factors DEPTH_BIN and ISLAND however, this seems like the standard adonis2 output provides but only using the data from the levels of the first factor DEPTH_BIN (it repeats this for all 3 pairwise combinations for levels of DEPTH_BIN but never for ISLAND and never for the crossed comparison of levels between factors - the latter is what i thought it does)???
$Shallow_vs_Middle
Permutation: free
Number of permutations: 999
Terms added sequentially (first to last)
Df SumsOfSqs MeanSqs F.Model R2 Pr(>F)
DEPTH_BIN 1 0.12942 0.129425 15.7494 0.10643 0.001 ***
ISLAND 4 0.12544 0.031360 3.8161 0.10315 0.002 **
DEPTH_BIN:ISLAND 4 0.04083 0.010208 1.2422 0.03358 0.256
Residuals 112 0.92039 0.008218 0.75685
Total 121 1.21608 1.00000
Please advise
Thank you
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package