search

biometry / bipartite

repository for the bipartite R-package for network analysis
37 stars 17 forks source link

NODF values should range from 0 to 1 #12

Closed sjdritz closed 4 years ago

sjdritz commented 4 years ago

It looks like the output from the networklevel() function reports values for NODF and weighted NODF greater than 1 when these values should range from [0, 1]. I believe the issue is that the function is reporting e+01 when it should be e-01

image

cdormann commented 4 years ago

I don't know where you got the idea from that NODF and weighted NODF should be within [0,1]. That is not correct: Almeida-Neto et al. (2008) in their table 2 show that NODF has values between 0 and 100. Also, networklevel "only" calls nestednodf from vegan, so you may want to direct your criticism to vegan's maintainer. Personally, I find the different scaling and orientation of nestedness measures unnecessarily confusing, as also noted by Almeida-Neto et al. 2008. One has to be extra careful to interpret them correctly. But I don't think I can simply start programming them "my way" in bipartite, when the definitions are different.

jarioksa commented 2 years ago

This is an old and stale issue, but my comment: I do think that scaling in [0,1] would be more sensible, but the original paper multiplied the numbers by 100 to give scaling [0,100]. Many people are used to parts per cent numbers instead of parts per one, and we just followed the publication (to avoid complaints of getting "wrong numbers" of [0,1] instead of the correct ones in [0,100]).

cdormann commented 2 years ago

Dear Jari,

I tried my best to address the issue you raised in the next release (2.17) of bipartite. I fully agree that this was messy and hope to have cleared it up a bit (and documented it properly).

Cheers, Carsten

On 22. Mar 2022, at 10:04, Jari Oksanen @.***> wrote:

This is an old and stale issue, but my comment: I do think that scaling in [0,1] would be more sensible, but the original paper multiplied the numbers by 100 to give scaling [0,100]. Many people are used to parts per cent numbers instead of parts per one, and we just followed the publication (to avoid complaints of getting "wrong numbers" of [0,1] instead of the correct ones in [0,100]).

— Reply to this email directly, view it on GitHub https://github.com/biometry/bipartite/issues/12#issuecomment-1074906160, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABZGVIVDBX3RPJWSYB3JI6LVBGELRANCNFSM4QTRRNIA. You are receiving this because you modified the open/close state.

— — — — — — Carsten Dormann

Professor of Biometry and Environmental System Analysis Faculty of Environment and Natural Resources University of Freiburg Tennenbacher Str. 4 79106 Freiburg

Tel.: 0049 (0)761 203-3750 Fax.: 0049 (0)761 203 3751 Email: @.***