search

lamho86 / phylolm

GNU General Public License v2.0
30 stars 12 forks source link

Alpha warning when performing phylostep/phyloglmstep #33

Closed estherdr closed 3 years ago

estherdr commented 4 years ago

For my project I used the phyloglmstep implementation presented by rvosa in the previous issue (https://github.com/naturalis/phyloglmstep/blob/master/R/phyloglmstep.R).

During this I got the following warning: "The estimate of ‘alpha’ (8.03841658764875) reached the upper bound (8.06867371201789). This may simply reflect a flat likelihood at large alpha values, meaning that the phylogenetic correlation is estimated to be negligible"

This warning seems to indicate that the phylogenetic component is not of major importance, which is not a huge problem in itself in my opinion. Is my opinion correct in your eyes or do you think I should avoid using results that come with this warning?

Thank you in advance!

lamho86 commented 4 years ago

Hello,

Your interpretation of this warning is correct. You don’t need to avoid using results that come with this warning.

best, Lam

On Jul 1, 2020, at 10:48 AM, estherdr notifications@github.com wrote:

For my project I used the phyloglmstep implementation presented by rvosa in the previous issue (https://github.com/naturalis/phyloglmstep/blob/master/R/phyloglmstep.R https://github.com/naturalis/phyloglmstep/blob/master/R/phyloglmstep.R).

During this I got the following warning: "The estimate of ‘alpha’ (8.03841658764875) reached the upper bound (8.06867371201789). This may simply reflect a flat likelihood at large alpha values, meaning that the phylogenetic correlation is estimated to be negligible"

This warning seems to indicate that the phylogenetic component is not of major importance, which is not a huge problem in itself in my opinion. Is my opinion correct in your eyes or do you think I should avoid using results that come with this warning?

Thank you in advance!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lamho86/phylolm/issues/33, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABW3HP5AT3EREYEUBENUJKTRZM5DFANCNFSM4ONN3BEQ.