glmmPQL and random effects

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

glmmPQL and random effects

dbahrd
Hello R users,

I am trying to run a model with a binary response variable (nesting
success: 0 failure, 1 success) and 8 fixed terms. Nesting success was
examined in 72 cases in 34 territories (TER) during a 6 study years.
Territories are nested within 14 patches (PATCH). I want to run a model
taking into account these nested factors and repeated observation. To do
this, I assume that the best option is to use glmmPQL from MASS package.
Am I wrong?

In glmmPQL, I have included the random terms as follow:
random=~1|YEAR/PATCH/TER, but I am unclear if this syntax is right for
this case (?).

I would greatly appreciate any help!

Regards,

Hugo

Hugo Robles
Department of Animal Biology
University of León (Spain)

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Reply | Threaded
Open this post in threaded view
|

Re: glmmPQL and random effects

Spencer Graves
          With repeated measures, you will want to use the "correlation"
argument in "glmmPQL".  For help with how to do that, you need to know
that "glmmPQL" calls "lme" repeatedly.  Therefore, if you have any
trouble using it, I suggest you consult Pinheiro and Bates (2000)
Mixed-Effects Models in S and S-PLUS (Springer).  Over the past five
years, I may have consulted this book more frequently than any other,
second only perhaps to Venables and Ripley (2002) Modern Applied
Statistics with S, 4th ed. (Springer).  If you don't have these books, I
suggest you consider purchasing them -- and / or recommending to your
university library that they obtain copies.  I see on
"http://www.unileon.es/servicios/unileoncat/acquire_web_spi.htm" that
they invite "Sugerencias de Adquisitión".

          If you can't get "glmmPQL" to work with your favorite model, I
suggest you try passing the problem directly to "lme", possibly with
random numbers for a response.  After I was confident I understood the
"lme" syntax, I'd then return to "glmmPQL".

          If you still have troubles, please submit another post -- after first
the posting guide! "www.R-project.org/posting-guide.html".  I believe
the people who more closely follow the procedure in the Posting Guide
are on average rewarded by a faster rate of new knowledge acquisition.
I have only anecdotes to support that conjecture, but anecdotes can be
valuable in the absence of more systematic data collection and analysis.

          Espero que estos commentos le ayude.

          Spencer Graves
       
[hidden email] wrote:

> Hello R users,
>
> I am trying to run a model with a binary response variable (nesting
> success: 0 failure, 1 success) and 8 fixed terms. Nesting success was
> examined in 72 cases in 34 territories (TER) during a 6 study years.
> Territories are nested within 14 patches (PATCH). I want to run a model
> taking into account these nested factors and repeated observation. To do
> this, I assume that the best option is to use glmmPQL from MASS package.
> Am I wrong?
>
> In glmmPQL, I have included the random terms as follow:
> random=~1|YEAR/PATCH/TER, but I am unclear if this syntax is right for
> this case (?).
>
> I would greatly appreciate any help!
>
> Regards,
>
> Hugo
>
> Hugo Robles
> Department of Animal Biology
> University of León (Spain)
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html