search

EcoClimLab / ForestGEO-tree-rings

Repository for analysis of tree-ring data from 10 globally distributed forests (Anderson-Teixeira et al., in press, Global Change Biology)
2 stars 2 forks source link

coauthor comments #101

Closed teixeirak closed 3 years ago

teixeirak commented 3 years ago

Raquel:

Maybe I’m missing something, but this looks to me that you don’t use a unique model to identify the climate sensitivity while accounting for size effects (the stated objective of the m.s.). Researchers tend to use similar approaches than this two-step process, but the difference is that most of them use linear approaches. e.g. Alfaro-Sanchez 2019. https://doi.org/10.1016/j.agrformet.2019.107630 Alfaro-Sanchez 2020. https://doi.org/10.1007/s13595-020-00949-x I recommend to include a more specific statement about the differences of this new method with previous approaches and to remove the word simultaneously from the objectives.

Which are the explanatory variables of these GAM models? So, instead of using a classical detrending method, you are using a GAM, right? Include the formula of the model. You may want to check out some of the papers that specifically look for differences among detrending methods e.g. Sullivan 2016 (attached to the email)

teixeirak commented 3 years ago

Raquel:

Are these models run by site? It is not recommended to use random effects for variables with less than five levels, here species in all the sites but two. This could be really problematic during the review process. I recommend to run these analyses at the species level.

teixeirak commented 3 years ago

Are these models run by site? It is not recommended to use random effects for variables with less than five levels, here species in all the sites but two. This could be really problematic during the review process. I recommend to run these analyses at the species level.

@ValentineHerr , do we have species as a random or fixed effect in the climwin step?

teixeirak commented 3 years ago

Maybe I’m missing something, but this looks to me that you don’t use a unique model to identify the climate sensitivity while accounting for size effects (the stated objective of the m.s.). Researchers tend to use similar approaches than this two-step process, but the difference is that most of them use linear approaches. e.g. Alfaro-Sanchez 2019. https://doi.org/10.1016/j.agrformet.2019.107630 Alfaro-Sanchez 2020. https://doi.org/10.1007/s13595-020-00949-x I recommend to include a more specific statement about the differences of this new method with previous approaches and to remove the word simultaneously from the objectives.

I think it's still fair to say "simultaneously", as climate + other factors are included together in the GLS. True that the candidate driver step is separate, and this needs to be clear throughout. I'm going to split this out as a separate issue.

ValentineHerr commented 3 years ago

@ValentineHerr , do we have species as a random or fixed effect in the climwin step?

It is a random effect if there is at least 3 species, it is ignored if there is less.

teixeirak commented 3 years ago

@ValentineHerr , do we have species as a random or fixed effect in the climwin step?

It is a random effect if there is at least 3 species, it is ignored if there is less.

Thanks. Really, it shouldn't matter as the mean of residuals for all species would be very close to zero, correct?

ValentineHerr commented 3 years ago

Yes.

ValentineHerr commented 3 years ago

(sorry, I didn't mean to close)

teixeirak commented 3 years ago

That's okay, I believe we're done with this.