Hi all,
I have received this comment from a reviewer that Shannon's diversity results from qiime2 are non-linear and must be exponentiated first before testing by generalized linear models. Do I need to use the exponential Shannon for GLM? Is it available in Qiime2 or I need to use another way?
Hi @m.abdallah,
Welcome to the 
forum!
I've personally never heard of this. I tend to find shannon is continuous and independent identically distributed data. (Although that distribution isn't always normal; it might be worth checking to see what your distribution looks like.)
Did they provide a citation or justification other than the statement? Have you looked at other papers running similar models to see what they do?
There are totally other types of microbiome data that need a log transform in microbiome data, so maybe check that?
Best,
Justine
PS The QIIME 2 implementation wraps this version of Shannon
1 Like
Hi @jwdebelius,
Thank you so much.
I attached my Shannon results.
The reviewer justified his request by these:
http://www.loujost.com/Statistics%20and%20Physics/Diversity%20and%20Similarity/EffectiveNumberOfSpecies.htm
AND
"Entropy and diversity" in Oikos (2006)
In the other papers, as far as I know, no one exponentiate Shannon results before modeling. I have also used a log transformed Shannon values for my models. I also ran model assumptions tests and no potential issues related to normality or collinearity were detected.
shannon_vector.qzv (369.8 KB)
Best,
Mohamed
Hi @m.abdallah,
I might argue around Hill Numbers (the wikipedia article might be a good place to start) or maybe that it's not commonly done. But, I think ultimately, if you need to adjust the values, you need to go outside QIIME.
Best,
Justine
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.