Closed Hansam closed 2 months ago
Which version of vegan you are using? The version in CRAN and in github have different kind of scores.
Hello :),
I'm using the CRAN version. Could you explain why the two versions have different scores? And should the scores from the github package match the Szoecs paper?
Thanks a lot, Hanna
Returning back after long delays: I had now time to have a look at the prc() function and compare it against other alternatives.
I asked about version, because CRAN versions since vegan 2.5-1 use wrong kind of scores. The github version was fixed in November 2022 and since then it is similar as before 2.5-1 release of vegan. So older prc (before 2.5-1) and github version give similar results, but in between the result diverged a bit.
About scaling: the numeric scaling of vegan results is not fixed, but it depends on the options you choose. Arguments scaling
and const
will change numeric values. However, the results will differ only by scaling, that is, a constant multiplier. The relative differences will be unchanged.
Hello,
I am looking for some help with an issue I encountered while reproducing the PRC from Szoecs 2015 (as by the tutorial in the Supplementary Material: https://www.researchgate.net/publication/272094723_Supplemental_Material_for_the_paper_Analysing_chemical-induced_changes_in_macroinvertebrate_communities_in_aquatic_mesocosm_experiments_A_comparison_of_methods).
I am using the pyrifos data set that is supplied by the vegan package and also in Canoco 5 to validate my code for the use in GLP studies (similar to this closed issue [https://github.com/vegandevs/vegan/issues/17].
I can reproduce the results with the exception of the canonical coefficients which are approximately 10x lower. As a result, when plotting the data the shape of the curves is the same, but not the values on the y-axis.
All the distances and results are proportional and the statistical tests on the PRC also match. However, as this exercise is to validate the code, I will need to convince the GLP auditors that the code and its result are reliable. As I cannot explain the 10 fold difference in the coefficients this will be difficult.
Any help/explanation is very much appreciated!
Thank you, Hanna