R help - Mar 2004 - Creating its own family

Hello,
I would like to create my own family for glm modelling

If we consider a matrix Y, I would like to model Yij/var(Yj) with an 
inverse variance link.
I have create my own family inspired by the negative.binomial of MASS:

#########################################################
myfamily=function (varcol)
{
     env <- new.env(parent = .GlobalEnv)

     assign(".varcol",varcol,envir=env)
     famname="myfamily"
     link="inv col var"
     linkfun=function(.varcol,mu){mu/(.varcol)}
     linkinv=function(.varcol,eta){eta*(.varcol)}
     variance <- function (mu) rep.int(1, length(mu))
     validmu <- function (mu) TRUE

     dev.resids <- function (y, mu, wt) wt * ((y - mu)^2)
     aic <- function (y, n, mu, wt, dev) sum(wt) * (log(dev/sum(wt) * 2 * 
pi) + 1) + 2
     mu.eta=function (eta) rep.int(1, length(eta))
     initialize <- expression({
     n <- rep.int(1, nobs)
     mustart <- y
     })

     environment(variance) <- environment(validmu) <- 
environment(dev.resids) <- environment(aic) <- env

     structure(list(family = famname, link = link, linkfun = linkfun,
         linkinv = linkinv, variance = variance, dev.resids = dev.resids,
         aic = aic, mu.eta = mu.eta, initialize = initialize,
         validmu = validmu), class = "family")
}
#####################################################

But it does not work when I try it

 > Y=matrix(runif(50),10,5)
 > glm(as.vector(t(Y))~x,family=myfamily(rep(apply(Y,2,var),10)))
Error in family$linkfun(mustart) : Argument "mu" is missing, with no
default
 >

Hope that someone could help me,

Thanks in advances,
Sincerely.

St?phane DRAY
--------------------------------------------------------------------------------------------------

D?partement des Sciences Biologiques
Universit? de Montr?al, C.P. 6128, succursale centre-ville
Montr?al, Qu?bec H3C 3J7, Canada

Tel : 514 343 6111 poste 1233
E-mail : stephane.dray at umontreal.ca
--------------------------------------------------------------------------------------------------

Web                                          http://www.steph280.freesurf.fr/

I think you want

linkfun <- function(mu){mu/varcol}
linkinv <- function(eta){eta*varcol)}

etc, that is a function of a single argument.  Also, you can use lexical 
scope to capture varcol, as in the above: it will be in the environment of 
linkfun/inv.  You did not set its environment as MASS does, and you do not 
need to.  (MASS could have been written differently if it did not share 
code with S.)

On Tue, 9 Mar 2004, Stephane DRAY wrote:
> Hello,
> I would like to create my own family for glm modelling
> 
> If we consider a matrix Y, I would like to model Yij/var(Yj) with an 
> inverse variance link.
> I have create my own family inspired by the negative.binomial of MASS:
> 
> #########################################################
> myfamily=function (varcol)
> {
>      env <- new.env(parent = .GlobalEnv)
> 
>      assign(".varcol",varcol,envir=env)
>      famname="myfamily"
>      link="inv col var"
>      linkfun=function(.varcol,mu){mu/(.varcol)}
>      linkinv=function(.varcol,eta){eta*(.varcol)}
>      variance <- function (mu) rep.int(1, length(mu))
>      validmu <- function (mu) TRUE
> 
>      dev.resids <- function (y, mu, wt) wt * ((y - mu)^2)
>      aic <- function (y, n, mu, wt, dev) sum(wt) * (log(dev/sum(wt) * 2
*
> pi) + 1) + 2
>      mu.eta=function (eta) rep.int(1, length(eta))
>      initialize <- expression({
>      n <- rep.int(1, nobs)
>      mustart <- y
>      })
> 
>      environment(variance) <- environment(validmu) <- 
> environment(dev.resids) <- environment(aic) <- env
> 
>      structure(list(family = famname, link = link, linkfun = linkfun,
>          linkinv = linkinv, variance = variance, dev.resids = dev.resids,
>          aic = aic, mu.eta = mu.eta, initialize = initialize,
>          validmu = validmu), class = "family")
> }
> #####################################################
> 
> But it does not work when I try it
> 
>  > Y=matrix(runif(50),10,5)
>  > glm(as.vector(t(Y))~x,family=myfamily(rep(apply(Y,2,var),10)))
> Error in family$linkfun(mustart) : Argument "mu" is missing, with
no default
>  >
> 
> Hope that someone could help me,
> 
> Thanks in advances,
> Sincerely.
> 
> St?phane DRAY
>
--------------------------------------------------------------------------------------------------
> 
> D?partement des Sciences Biologiques
> Universit? de Montr?al, C.P. 6128, succursale centre-ville
> Montr?al, Qu?bec H3C 3J7, Canada
> 
> Tel : 514 343 6111 poste 1233
> E-mail : stephane.dray at umontreal.ca
>
--------------------------------------------------------------------------------------------------
> 
> Web                                         
http://www.steph280.freesurf.fr/
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
> 
> 
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595