rarefying vs. rarefaction

[Geoff Cook]

Hi,

I recently read the McMurdie and Holmes paper (2013) that reported why rarefying microbiome data is inadmissable. On p.3, the authors state  that "Rarefying is now an exceedingly common precursor to microbiome multivariate workflows that seek to relate sample covariates to sample-wise distance matrices; for example, integrated as a recommended option in Qiime's 'beta_diversity_through_plots.py' workflow...."

I also found confirmation on the qiime forum that the alpha and beta diversity workflows "rarefy otu tables as part of the steps they perform" (see here https://groups.google.com/forum/#!msg/qiime-forum/PqFk9vhyiL4/rWl5-6QFtmYJ.

However, the results from alpha_rarefaction.py appear to align with Sanders (1968) and Gotelli (2001). McMurdie and Holmes (2013) outline a clear difference between "rarefy," a a normalization technique, and "rarefaction," a technique for species richness estimation (i.e., taxon resampling curves). 

Given McMurdie and Holmes offer Qiime's "beta_diversity_through_plots.py" as a workflow that rarefies (i.e., normalizes) the otu table prior to analysis, I am a little confused about what, exactly, is being done to the otu table. Is it being rarefied, as per McMurdie and Holmes, or is a rarefaction, as per Sanders and Gotelli, being performed?

Thanks,

Geoff 

[justink]

Without being familiar with Sanders (1968) and Gotelli (2001), I'd say beta_diversity_through_plots and similar objects in qiime match  "rarefy, a normalization technique". Throughout most of the qiime documentation, 'rarefy' and 'rarefaction' are used usually as the verb and the noun representing the same concept, and that concept is sampling counts from an OTU table without replacement, for a variety of purposes including species richness estimation.

Best!