I have data on coral juveniles density along with live coral cover, hard substratum cover and macroalgal cover from 6 study sites categorized in to 3 groups based on anthropogenic pressure viz. severe, moderate and low. I want to explore the extent to which the coral juvenile density is associated/influenced with/by the live coral, hard substratum and macroalgal cover. I tried multiple regression analysis. But my data fails to meet the assumptions of multiple regression analysis that there exists a negative correlation/no correlation between the juvenile coral density and other 3 parameters individually.Could anyone suggest me the best statistical method to explore it?
Helio Clemente Jose Cuve @ Thanks for your kind reply. I tried Pearson and Spearman correlation. I get negative R values.
Dear Manikandan,
sound s like you should get into generalised linear modelling (GLM) or in case you expect non-linear relationships between your influencing variables (site, coral cover, hard substratum cover and macroalgae cover) and and the dependent variable (juvenile coral density), generalised additive modelling (GAM). Both model types can be well implemented with R and there is a lot of literature and application examples on the web.
Here two potential books that may help you:
Crawley, M. J. 2005. Statistics - An introduction using R, John Wiley & Sons, Chichester.
Zuur, A., F., Ieno, E. N., Walker, N. J., Saveliev, A. A., and Smith, G. M. 2009. Mixed effects models and extensions in ecology with R, Springer, New York. 574 pp.
In case some of your predictors show covariaton, you may have to remove one of them from your model.
Alternatively to GLM/GAM you could try principal component or factor analysis. These methods help you to look at which variables are correlated by checking, which variable is "loading" onto a sort of virtual axis (=component or factor). PCA/Fa to my knowledge, however, only work with continious variables, so you may get problems acoounting for the influence of site.
Hi Manikandan,
It may be possible using some of the functions in PRIMER which is a multi-variate statistical package. Within this software you can use a function called BIOENV which will explore your biological data (corals) and test them against your environmental variables. The 'BEST' function in the package may also be able to give you information on the variable or combination of variables that best explain any relationships between environment and biology. Without knowing more about your project it is difficult to be sure but it might be worth a look.
See CLARKE KR, GORLEY RN (2001, 2006, 2015) PRIMER v5, v6 & v7: User manual/tutorial, PRIMER-E, Plymouth UK, 91pp, 192pp & 296pp [5,231]
PRIMERe is available here http://www.primer-e.com/
Other packages are available and you should look at possible using a Canoical Correspondance Analysis or as Wolfgang suggests a PCA, but I would look at PRIMER first as it is written with biologists in mind and the software developers are easily contacted and provide great support.
Best wishes,
Kev
Wolfgang @ Thanks for your valuable suggestion. I will try both GLM and GAM and look out for the best one. I will get back to you for any doubts. Since I'm unfamiliar with R software, I have to start reading from the basics to be more familiar with it.
Nicolas@ Thanks for the suggestion. My data is not a continuous variable, Instead it is spatial data. I fear whether my data will meet the assumptions of PERMANOVA. Still will explore the possibilities. Thank you Once again
Kevin @ Thanks for your suggestion. I have used other univariate and multivariate analysis in PRIMER to discriminate the study sites based on the data I have. I will try BIOENV..and will get back to you.
There is also a PERMANOVA+ package in the most recent version of PRIMER.
Kev
- Badges
- Science topic