confidence level in pairwise comparison

[cres...@gmail.com]

Hi there,

First of all, thanks a lot for this useful package and the user-friendly tutorials!

I have a question regarding pairwise comparisons: Is there an accepted level of confidence for probabilities above which I can state that the distributions are actually different between my two species ? 75 %? 95 % ?

And I have a comment regarding potential future developments. Would it be possible to include different TEFs in the same model? I have teleosts and sharks in my assemblage. They are fueled by the same pelagic and benthic production, but I apply different TEFs. I have used two separate models that I merged afterwards to plot and test the distributions but I could be helpful (if possible!) to consider that. This could be useful when working on the whole assemblage (eg calculating average TP and alpha) rather than on the values for each species!

Thanks again anyway!

Cheers

Pierre

[Claudio Quezada Romegialli]

Hi Pierre!

Thank you very much for your comments ;)

Regarding your question, as we are dealing with Bayesian statistics instead of the frequentist approach, there is no 5% (p < 0.05) rule that we can follow. If you are interested in saying that two distributions are different, then I would use at least a 95 % threshold (even 99 % if you want to be more confident). Also, there are 2 things to mention: the function compareTwoDistributions (or pairwiseComparisons) uses 2 algorithms to compare 2 distributions. If you use the Bhattacharyya coefficient (test = "bhatt") it will measure the overlap between 2 distributions, while if you use test = "<", "<=", ">" or ">=", it will randomly take 2 values from each distribution and compare them using the logical test (this is the approach that Andrew Jackson used in SIBER).

Regarding your comment to future development, it would be possible to use different TEF in the same model, but I guess it is easier to separate the data frame as you did, calculate the model with different TEFs per separate, and then combine the output. This should be pretty straightforward though. If you want I can prepare a vignette on this, because it will take a few lines to do it! If you already publish your data, I can use the same data with the vignette.

Cheers

Claudio

--
To download tRophicPosition package visit https://github.com/clquezada/tRophicPosition
---
You received this message because you are subscribed to the Google Groups "tRophicPosition" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trophicposition-s...@googlegroups.com.
To post to this group, send email to trophicposi...@googlegroups.com.
Visit this group at https://groups.google.com/group/trophicposition-support.
To view this discussion on the web, visit https://groups.google.com/d/msgid/trophicposition-support/531174d9-b71a-48ce-8742-0382788fc86d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Dr. Claudio Quezada-Romegialli

Profesor Asociado

Departamento de Biología

Facultad de Ciencias Naturales y Exactas

Universidad de Playa Ancha

Móvil: +56 9 9038 0148

Teléfono: +56 32 250 0519