Background- I am studying skin bacterial changes in salamanders before and after translocation and have two source populations. I am trying to look at differences in skin bacterial OTUs between the two source streams prior to translocation. I have repeat samples of individuals so I used that strata function to constrain permutations based on individual. However, when I run the permanova, I get a p-value of 1 for both my Jaccard and Bray Curtis dissimilairty matrices which seems really weird to me. Anyone have any ideas why this might be happening? Is it something to do with the strata function? I can provide more data or details if need be. Example below:
D4.1000.hb.pre - the distance matrix
pre.source - source stream
strata = individual
* Bray Curtis formula = adonis(D4.1000.hb.pre ~ pre.source, strata = beta.hb.pre$Animal.ID, data=beta.hb.pre, permutations = 1000)
Blocks: strata
Permutation: free
Number of permutations: 1000
Terms added sequentially (first to last)
Df SumsOfSqs MeanSqs F.Model R2 Pr(>F)
pre.source 1 1.6274 1.62740 4.5042 0.09686 1
Residuals 42 15.1750 0.36131 0.90314
Total 43 16.8024 1.00000
* Jaccard formulay = adonis(D5.1000.hb.pre ~ pre.source, strata = beta.hb.pre$Animal.ID, data=beta.hb.pre, permutations = 1000)
Blocks: strata
Permutation: free
Number of permutations: 1000
Terms added sequentially (first to last)
Df SumsOfSqs MeanSqs F.Model R2 Pr(>F)
pre.source 1 1.0961 1.09610 2.6986 0.06037 1
Residuals 42 17.0594 0.40618 0.93963
Total 43 18.1555 1.00000
Because you have a very significant difference. Just trust the result!
Dear Emilly,
In my understanding, you want to test effect of the source stream, but your settings for restricted permutation are permuting values only within the individuals defined by your strata, but not across the levels of your explanatory variables (i.e. source stream). Therefore the resulting p-value is 1.
What you probably want to do, is to treat the individuals as blocks which should be freely permuted between the source streams. Such settings can be easily done with permutations = how() argument in vegan.
Best regards,
martin
Yes, the set-up doesn't look right, and p=1 says that there is absolutely no difference
- Badges
- Science method