So my question is if there are any analyses that would compare two groups diversity that have been measured differently. The groups that I want to compare are representatives of the past (historical) genera diversity and the present community genera diversity. The Past community diversity is measured and calculated on biomass whereas the Present data is calculated with presence/absence measurements.
I was mainly concerned to as if I could analyse and compare these two data-sets statistically, and if I cam ,which test should I use? So far, I've only compared the diversity with visual methods, such as pie-charts, to present the percentage of a specific diversity. But can I use any analysis to see which genera is contributing with the decrease/increase of diversity and which genera is loosing/wining the most?
Thanks!
Dear Stefan,
Your query is not very clear. However, you should probably consider using Correspondence Analysis (CA). Be aware, though, that you should not use "tests". Rather, you should find associations, which CA will provide. It (essentially) will reduce the number of dimensions you need to clarify associations.
You must convert one of your data axes to categorical variables. The other, from what I gather after reading your text, is already categorical.
One further caveat: if you have a large number of species and modern variable assessments, then your contingency matrix (and the correspondence matrix) is HUGE. You might have to program CA (using SVD of the correspondence matrix) directly, rather than using the routines available in software packages, because they might not be able to handle such large matrices and, to wit, might also not be able to doe SVD in extended precision arithmetic.
Hope this helps, Hermann
I probably would have converted the data set (measured on biomass) into factorial scale (say 0-1) and then performed unconstrained ordination analysis, as suggested by Hermann Prossinger. Procrustes Randomization Test of both the matrices (scaled on 0-1) may be one possibility to test the degree of concordance between both the matrices. Then I would go for Detrended Correspondence Analysis (DCA) or CA with the combined data set to approximate spatial and temporal gradients, and then would extract sample scores of the axis representing the temporal gradient. I think, these axis scores can be used to find out which genus has increased or decreased to has remained more or less unchanged over time period.
Species richness is related to biomass, so it is valid to compare the two sets of data. See Journal of Ecology, Vol. 84, No. 2 (Apr., 1996), pp. 293–295. You can do the comparison statistically (e.g. using Simpson's diversity index) or, as Hermann stated, using associations.
Hi,
I would use Hill numbers (see Jost 2006 in Oikos and 2007 Ecology), first you need to calculate the efective number of species and using a bootstrap you can compare your samples with no problems, you can do this with package vegetarian in R
Hi Stefan,
The first thing I would suggest is have a read of "Measuring Biological Diversity" by Anne Magurran (especially the chapter on commonness and rarity of species and B diversity).
Then I would look closely at the manner in which your data was collected. Are the means comparable or are you dealing with two different scales of measure? Can you safely convert the historical data to presence/absence with the same level of confidence as the recent data? Can you calculate biomass with the current data set?
Once you have answered these sorts of question, comparisons of diversity measures can be achieved using your eyes (graphical representation of data is always the best first step in analysis), Null Models, ABC curves (Warwick 986), Rarefaction, etc. Then there are the Beta tests for comparing changes in diversity between the two data sets (a must do, given the temporal factor). If you have environmental measures, then I would suggest a reason for heading down the path of ordination (CCA) to investigate what may have lead to any changes in species composition (if any).
maybe I'll try to standardize the measurements of diversity in each groups (if your data allow it): in this manner you'll have mean diversity=0 and standard deviation=1 in both groups (hystorical and modern). Then you can analyze the influence of different variables (such as genera abundance or geographycal feature) on diversity and if this influence varies in the two groups. As regard the direct comparison of diversity between the two groups... it seems quite difficult: the book suggested by Eddy Cannella would be very useful.
I agree that the Magurran book suggested by Eddy Canella would be a useful start. I also add that whenever I am faced with this type of data I start with a Monte Carlo simulation of the sampling process to give me some indications of the variability and the likely distribution of the relevant means and variances.
Dear Hermann Prossinger,
Thank you for your advice on the CA method to cut down on the multiple dimensions. My current species assemblage has a maximum of 87 species but I’m mainly working on genus level (around 30 genera) so that will also reduce the matrix size. I’m also working with less than 10 ecological trait variables so the matrix should not exceed any limitations, and R seems to be doing a good job so far with even bigger matrices that I have been working on before.
Thank you Kuber P Bhatta,
I haven’t thought about converting my data into a factor scale, this is a great idea! A concordance between the two matrices sounds also like a good idea. I’m not sure however about the temporal gradient as we will not have determined the age of the historic samples, we can just say that they are representing a historic community.
Andrew Paul McKenzie Pegman,
Thank you for the article, I’ll definitely check it out and discuss it with my supervisor.
Pedro Luna,
Awesome, thank you very much. I’ll check it out and try to use it for my data-set.
Hi Eddy Cannella,
The questions you asked are valuable and I will definitely have to answer these to my manuscript! Sadly I cannot calculate biomass from the current data-sets as these are just observations from divers when they were mapping the coral community diversity and distribution. ABC curves are a no-can-do because of the lacking biomass from the present data. The Rarefactions curves showed some interesting results and are currently being discussed with a visual comparison. I’m trying to use ecological trait values and possibly present environmental data values for the CCA to show which environmental variable may have controlled to the communal diversity and distribution.
Thank you Laura Stancampiano,
I’ll experiment with this method to see what kind of results it will produce.
Hubertus Jacobus Van Hensbergen,
The Monte Carlo simulations is completely new to me. I’ll look to see what it’ll show and produce when I test it. We have not been taught anything about this simulation in our stat-courses. Thank you for the idea!
Thank you all for your contributions and all your advice has been greatly appreciated! I’ll come back to this post if I stumble upon any further questions regarding the ideas and methodologies you all suggested.
Thank you again
Sincerely,
Stefan Husa
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology