R help - Feb 2014 - package mgcv - predict with bam: Error in X[ind, ] :, subscript out of bounds

Dear Simon, your note below says "bs="re" specifies a Gaussian
random
effect ". I have been using bs = "re" for data modeled with
Poisson and
binomial distributions, or variants thereof (e.g., quasi-Poisson). Have 
I erred in assuming bs ="re" can be used to obtain random effects for 
such data? Will Shadish

- Actually this is ok. mgcv exploits the duality between quadratically
penalized smooths and Gaussian random effects to allow random effects to
be specified this way. bs="re" specifies a Gaussian random effect with
corresponding model matrix given by model.matrix(~site-1). (More
generally s(x,y,z,bs="re") specifies a gaussian random effect with
model
matrix given by model.matrix(~x:y:z-1), with obvious generalization to
more or fewer variables). See mgcv help file ?random.effects for more.

best,
Simon


-- 
William R. Shadish
Distinguished Professor
Founding Faculty

Mailing Address:
William R. Shadish
University of California
School of Social Sciences, Humanities and Arts
5200 North Lake Rd
Merced CA  95343

Physical/Delivery Address:
University of California Merced
ATTN: William Shadish
School of Social Sciences, Humanities and Arts
Facilities Services Building A
5200 North Lake Rd.
Merced, CA 95343

209-228-4372 voice
209-228-4007 fax (communal fax: be sure to include cover sheet)
wshadish at ucmerced.edu
http://faculty.ucmerced.edu/wshadish/index.htm
http://psychology.ucmerced.edu

The two distributions are different. The random effect is assumed to
be a Gaussian random variable, just as it is with the GLMMs in the
lme4 package. It is fine to use such a random effect within a GAM with
a non-Gaussian error distribution, like the ones you describe using.

HTH

Gavin

On 3 February 2014 15:00, William Shadish <wshadish at ucmerced.edu>
wrote:
> Dear Simon, your note below says "bs="re" specifies a
Gaussian random effect
> ". I have been using bs = "re" for data modeled with Poisson
and binomial
> distributions, or variants thereof (e.g., quasi-Poisson). Have I erred in
> assuming bs ="re" can be used to obtain random effects for such
data? Will
> Shadish
>
> - Actually this is ok. mgcv exploits the duality between quadratically
> penalized smooths and Gaussian random effects to allow random effects to
> be specified this way. bs="re" specifies a Gaussian random effect
with
> corresponding model matrix given by model.matrix(~site-1). (More
> generally s(x,y,z,bs="re") specifies a gaussian random effect
with model
> matrix given by model.matrix(~x:y:z-1), with obvious generalization to
> more or fewer variables). See mgcv help file ?random.effects for more.
>
> best,
> Simon
>
>
> --
> William R. Shadish
> Distinguished Professor
> Founding Faculty
>
> Mailing Address:
> William R. Shadish
> University of California
> School of Social Sciences, Humanities and Arts
> 5200 North Lake Rd
> Merced CA  95343
>
> Physical/Delivery Address:
> University of California Merced
> ATTN: William Shadish
> School of Social Sciences, Humanities and Arts
> Facilities Services Building A
> 5200 North Lake Rd.
> Merced, CA 95343
>
> 209-228-4372 voice
> 209-228-4007 fax (communal fax: be sure to include cover sheet)
> wshadish at ucmerced.edu
> http://faculty.ucmerced.edu/wshadish/index.htm
> http://psychology.ucmerced.edu
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Gavin Simpson, PhD