Hi Sileesh,

I agree with Malcolm, your question is way too broad. If you are looking for specific statistical analysis suitable for your data set you need to ask more specific questions. Stating what hypothesis you are trying to test would be a good start. However, there is a very good statistical package for this type of analysis called Primer-E that I think you will find very useful. This is the link:

http://www.primer-e.com/

I have used this package in the past and it is well worth looking at. On their web site you will find a section on "Papers" and if you go to it you will find a lot of publications that have used the package and statistical analysis that are most likely suitable for your needs as well. . Many of these papers will give you a very good idea what you need to do. This is the link:

http://www.primer-e.com/Primary_papers.htm

I am attaching a must read paper by Field et al (1982) that will clearly explain some of the statistical approaches for analysing multispecies distribution patterns. I have also attached papers by Clarke and Green (1988) and Clark and Ainsworth (1993) that I consider a must reads that should give you more ideas. The paper by Potter et al (2001) looks at the long-term changes in fish abundance and species composition so it may fit your needs as well. If you are interested in food webs than the paper by French et al (2013) should be of interest to you. If you are dealing with conservation issues then the paper by Purcell et al 2014) should help. Finally the paper by Vialle et al (2014) should give you some ideas on how to analyse the relationships between species numbers and environmental conditions. Go to “Downloads” and have a look at the manuals, these should also be very helpful, even if you do not have the program. I hope this helps.

Tomas

Hi Sileesh Mullasseri

It depends on your aims and then also the kind of data you have and the type of relationships the fish abundances might have with your environmental data. Assuming it's exploratory analysis, I'd probably start with something like non-metric multidimensional scaling (nMDS). It's the least prone to giving misleading relationships of the commonly used multivariate models which is great, but the fitting of vectors of environmental change onto an ordination is just correlative (that might still be OK).

Best wishes

Malcolm

Hi Sileesh,

I agree with Malcolm, your question is way too broad. If you are looking for specific statistical analysis suitable for your data set you need to ask more specific questions. Stating what hypothesis you are trying to test would be a good start. However, there is a very good statistical package for this type of analysis called Primer-E that I think you will find very useful. This is the link:

http://www.primer-e.com/

I have used this package in the past and it is well worth looking at. On their web site you will find a section on "Papers" and if you go to it you will find a lot of publications that have used the package and statistical analysis that are most likely suitable for your needs as well. . Many of these papers will give you a very good idea what you need to do. This is the link:

http://www.primer-e.com/Primary_papers.htm

I am attaching a must read paper by Field et al (1982) that will clearly explain some of the statistical approaches for analysing multispecies distribution patterns. I have also attached papers by Clarke and Green (1988) and Clark and Ainsworth (1993) that I consider a must reads that should give you more ideas. The paper by Potter et al (2001) looks at the long-term changes in fish abundance and species composition so it may fit your needs as well. If you are interested in food webs than the paper by French et al (2013) should be of interest to you. If you are dealing with conservation issues then the paper by Purcell et al 2014) should help. Finally the paper by Vialle et al (2014) should give you some ideas on how to analyse the relationships between species numbers and environmental conditions. Go to “Downloads” and have a look at the manuals, these should also be very helpful, even if you do not have the program. I hope this helps.

Tomas

Hi Sileeshi,

If you are interested in a form of community structure analysis where you have multivariate compositional data (abundance data of multiple fish species - potentially over multiple sites/stations). Then Tomas have provided you with sufficient information to get you started using Primer-E software which is proprietary (meaning you have to pay for it). If you are familiar with R, freely available and widely used environment for statistics and more, there are range of libraries (packages) that allow you to do all your community structure related analysis and more. e.g. libraries to check vegan, labdsv but there are more.

Best Regards

Hi Silesh,

Canonical Correspondence Analysis (CCA) is a best statistical analysis in environmental parameters relation to species assemblages. Pls try this is best statistical method.

Remember you will need to normalise your data before running any model to test a relationship between biotic and abiotic factors.

In addition to all good previous answers, I would suggest a practical guide for biodiversity analysis with R. https://www.worldagroforestry.org/downloads/Publications/PDFS/b13695.pdf

It does not cover a comprehensive theoretical background but you can find a lot of examples with codes.

You may also want to look into non-parametric methods. George Sugihara's lab has some nice techniques. See http://scrippsscholars.ucsd.edu/gsugihara/publications

For this method you do need a decently long time series, and this is often a limitation. Additionally, for this method DO NOT normalise the data to start with.

Dear all,

I agree with all the previous answers. However, I have some valuable suggestions from an experienced statistical scientist. His advice is as follows,

As regards the type of data analysis relating species diversity with environmental factors there are two sets of possibilities.

If the normal assumptions are held valid then,

1. Try canonical correlation between the two sets of data.

2. Redundancy analysis also can relate multivariate response with multiple factors.

If the normality assumptions don't hold good then opt for PRIMER software based solutions. Especially BIOENV analysis based on dissimilarity matrices is the best one. This is also available in the vegan package of R.

Hope the answer is useful.

You can do it by either using Primer e software or R. I would like to do it with something like non-metric multidimensional scaling (nMDS) .