Time-series data and fixed levels of similarity in Primer

[vanguard605]

Hi,

Apologies if this is not posted in the right section. I am currently in the process of analyzing a time-series dataset where the same biological communities are routinely sampled annually. When I depict the structure of the communities, I like to use some hierarchical clustering. I try to group the communities sampled from a particular year into 3-4 clusters of at least three communities.

Since the data comes from the same region, in the same season, is it unreasonable to expect to get some fixed level of similarity among years, so that the communities fall into the same number of clusters, assuming no undue influence from some extreme exogenous event?

Please know that I work with these datasets with the Primer software context as I have to be to swap results with collaborators.

All responses are duly appreciated.

 

[bugman]

Hi,

Since the data comes from the same region, in the same season, is it unreasonable to expect to get some fixed level of similarity among years, so that the communities fall into the same number of clusters, assuming no undue influence from some extreme exogenous event?

All responses are duly appreciated.

No it its not an unreasonable assumption. But this will depend on the range of your critters and at what scale you are sampling at. I would need a bit more information about your biological data before I can comment further.

Is this all you are asking? Or are you having trouble with Primer-E itself?

 

[vanguard605]

No, I am having a problem with Primer. More specifically, I work with family-level abundance data of communities from sites along a transect. These sites are visited yearly. My general approach is to group each year's data individually into groups of 3-4 clusters consisting of at least 3 communities. The level of similarity I choose is specific to each year's data, but I find them to be quite similar. This has led me to wonder if I can find a fixed-level of similarity that can be readily applied to every year. Does this involve playing around with each year's data? This would be difficult since I have about 10 years worth of it.

 

[bugman]

No,you can't pre-define cutpoints but you can use the simprof routine to test whether your groups are different or not based on a specific alpha value. This may not be available in versions earlier than V6.

 

[bugman]

The level of similarity I choose is specific to each year's data, but I find them to be quite similar.

You have lost me here. I dont understand what you are doing!