[R-sig-ME] Fwd: RE: static vs dynamic factors together?

lmpitombo lmpitombo at ufscar.br
Fri Oct 9 20:02:53 CEST 2015


Dear Dr. Bolker,

When I specify the fixed effects I usually include both treatments and 
covariates in the same analysis - I´m looking for which factors better 
explain y. For instance, I have treatments 0, 50, 75 and 100 and the 
covariate "alpha". The treatments explain my variable much more often 
than the covariate.

So, the question that arises is if it is right give the same weight for 
treatments and covariates... Should be included some penalty for 
treatments?


Thanks again,
---
      Leonardo M. Pitombo
       UFSCar - Sorocaba
          Lab. de Solos

-------- Mensagem original --------
Assunto: RE: static vs dynamic factors together?
Data: 2015-10-08 23:05
De: "Bolker, Benjamin" <bolkerb at mcmaster.ca>
Para: lmpitombo <lmpitombo at ufscar.br>

   I don't really understand the question.  It might be better for
r-sig-mixed-models at r-project.org ... As far as I understand what
you're asking, I don't see any conceptual difference between static and
time-varying predictor variables ...

   cheers
     Ben Bolker

________________________________________
 From: lmpitombo [lmpitombo at ufscar.br]
Sent: Wednesday, October 07, 2015 9:17 AM
To: bolker at mcmaster.ca
Subject: static vs dynamic factors together?

Dear Dr. Bolker,

You always have payed attention in my questions in the "github" about
practival issues. Now, I have submmited a manuscript and one reviewer
asked something that I imagine I might reply but I would like your
(informal) input/ advice.

Here we go:

I have treatments (static factors) and environmental parameters (dynamic
factors)

To test which factors better explain the variable in test, usually I put
all together. Is it right, once the dynamic factors are associated with
a variance and the static parameters not? Is there any penalty in the
models for the static parameters?

Best Regards,
--
       Leonardo M. Pitombo
        UFSCar - Sorocaba
          Lab. de Solos
         (15) 3229 8842



More information about the R-sig-mixed-models mailing list