I am currently using Hmsc package to apply JSDMs to freshwater fish and mussel communities. Most of all, I want to study the correlations between the species. Is it correct to identify a correlation as positive or negative with at least 85% posterior probability? Or should I increase this to 90/95%?
Correlation Coefficient = +1: A perfect positive relationship. Correlation Coefficient = 0.8: A fairly strong positive relationship. Correlation Coefficient = 0.6: A moderate positive relationship. However, the two most commonly used statistical tests for establishing relationship between variables are correlation and p-value. Correlation is a way to test if two variables have any kind of relationship, whereas p-value tells us if the result of an experiment is statistically significant. So don't forget the p value too :)
This piece by Poggiato et al. (2021) TREE (Article On the Interpretations of Joint Modeling in Community Ecology
) can help you with the interpretation of your results.
I am interested in this because I am also using this package (and modeling freshwater mussels)!
I'm still learning about this package, but from reading examples, authors of the package provide a default supportLevel of 0.95. If you only see correlations with 0.85 then I'm not sure what the interpretation would be?
It looks like it's been over a month since you asked this question - curious to know if you found anything out since then.
What you mention as the 'posterior probability' comes from the confidence level of the credible interval (the bayesian counterpart of the confidence interval, I won't dig into it but see e.g. https://towardsdatascience.com/statistics-101-credible-vs-confidence-interval-af7b7e8fdd79) of the correlation term. Depending on the confidence level you set, you obtain a given credible interval, which can either overlaps zero or not. If it overlaps zero, than the parameter is not statistically significant at that confidence level (as it is your case at 95%). Of course, when you choose a less conservative confidence level (e.g. 85% in your case), then the credible intervals are narrower, and you will find more significant correlations.
As such, there's no 'correct' or 'incorrect' way of doing, but just a more or less conservative confidence level that you accept.
Concerning the ecological interpretation of your results, see Article On the Interpretations of Joint Modeling in Community Ecology
as suggested above.
Thank you all for the answers. Giovanni Poggiato clearly answers what I was trying to solve.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis