search

qiime2 / q2-diversity

BSD 3-Clause "New" or "Revised" License
4 stars 45 forks source link

Canberra Adkins form #221

Closed wasade closed 6 years ago

wasade commented 6 years ago

This pull request introduces Canberra-Adkins, which is a normalized variant of the Canberra metric. This metric is more similar, although not identical, to the "Canberra" metric of QIIME1. The slight difference stems from a bug in the QIIME1 implementation in which the number of nonzero elements used in the scaling factor was over estimated.

Co-authors on this include: @gregcaporaso @antgonza @ElDeveloper @rob-knight and @justin212k following internal discussion. The discrepancy with QIIME1 and QIIME2 was first identified by Dr. Alexey Melnik.

TODO:

thermokarst commented 6 years ago

note to self - before merging remember to add in the co-author tags for the aforementioned authors.

wasade commented 6 years ago

Ideally it would go into scipy but we decided it would be a lot faster to get it working here than to try and add to scipy. skbio is certainly an intermediate option.

ebolyen commented 6 years ago

Yeah citations don't quite work for these amalgamated methods. Each of these should be an action on its own and then beta can become a pipeline. Until then, there's not much we can do about the citations.

On Tue, Nov 6, 2018, 10:26 AM Daniel McDonald <notifications@github.com wrote:

@wasade commented on this pull request.

In q2_diversity/_beta/_method.py https://github.com/qiime2/q2-diversity/pull/221#discussion_r230952206:

@@ -124,9 +125,24 @@ def beta(table: biom.Table, metric: str, def aitchison(x, y, **kwds): return euclidean(clr(x), clr(y))

  • def canberra_adkins(x, y, **kwds):

  • Table 1 of Faith et al 1987 Plant Ecology 69: 57-68

Right, but this citation is only relevant when a specific metric is used not in general for the method?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/qiime2/q2-diversity/pull/221#discussion_r230952206, or mute the thread https://github.com/notifications/unsubscribe-auth/ADyuZBmk6QzTUVYLArH_EWkyV3ctFnm7ks5usMkagaJpZM4W_wQB .