Hi Ara,

It seems that they obtained one value of Bray-Curtis dissimilarity between each site (Y) that they plotted against the euclidean distance between each site (X, in km). Indeed, the vegdist function returns a "score for each pair of sites" more than a "pairwise scores for each site".

So in other words, they didn't get one single dissimilarity score for a sample site but one dissimilarity score for each pair of sites. By the way, it is possible to obtain a single dissimilarity score for a given site  for instance by averaging all the dissimilarity scores obtained between this site and all the other sample sites, thus providing an overall dissimilarity score of the considered site compared to all the other sites. Then, this new 'metric' can be plotted in function of environmental variables for example to investigate dissimilarity patterns within environmental gradients. However, this approach tends to strongly "smooth" the signal, but this is another debate.

Hope it helps !

All the best,

Anthony

Hi Ara,

It seems that they obtained one value of Bray-Curtis dissimilarity between each site (Y) that they plotted against the euclidean distance between each site (X, in km). Indeed, the vegdist function returns a "score for each pair of sites" more than a "pairwise scores for each site".

So in other words, they didn't get one single dissimilarity score for a sample site but one dissimilarity score for each pair of sites. By the way, it is possible to obtain a single dissimilarity score for a given site  for instance by averaging all the dissimilarity scores obtained between this site and all the other sample sites, thus providing an overall dissimilarity score of the considered site compared to all the other sites. Then, this new 'metric' can be plotted in function of environmental variables for example to investigate dissimilarity patterns within environmental gradients. However, this approach tends to strongly "smooth" the signal, but this is another debate.

Hope it helps !

All the best,

Anthony

Anthony,

That is perfect. Thank you. I was having a difficult time understanding how that plot was made.

ara

I want to check my understanding here. I can approach this problem in two ways. 

One is to average the scores for the "pair of sites" but this might smooth out any signal. This I understand. data.dist object will return all the distances for a pair of sites. I can write some R code to wrangle this into an average. 

The other way is (in the paper) was to get one value of Bray-Curtis dissimilarity between each site. I am confused on how this works. Rather, what the workflow/R script looks like to make that happen. 

Thanks again for the help and putting the time in.

ara

Hi Ara,

I think you understand it well, so I don't really see why you are confused regarding the analyses in the paper.

The two approaches (we can call them "averaging" and "distance") have two different goals. The "distance" one test if the dissimilarity between sites increases or decreases when the distance between sites increases (to distinguish between local and global effects for instance). On the other hand, by averaging these scores by site, you try to sum up the information into a 'general' index to test if the sites are more or less different compared to all the other sites depending on environmental variables such as air temperature or forest cover. What's important here is that you cannot really swap the two approaches. You cannot use distance vs. the averaged index. Similarly, you can only use the 'distance' approach to link dissimilarity scores to differences between sites, such as spatial distance, difference in latitude, difference in temperature, etc.

I don't know how to be clearer, but this is two distinct approaches. If your data make it possible, I think that the 'distance' approach worth the try first.

All the best !

Anthony

Ok, thanks for the clarifying!

ara

I've just seen a brand new article that could very probably interest you:

Socolar JB, Gilroy JJ, Kunin WE, Edwards DP. 2016. How should beta-diversity inform biodiversity conservation? Trends in Ecology & Evolution 31: 67-80.

You can find inside other examples of what we have been discussing.

Anthony

http://www.sciencedirect.com/science/article/pii/S016953471500289X

Article How Should Beta-Diversity Inform Biodiversity Conservation?