R devel - Jul 2009 - How to use formulas for multivariate clustered data?

Dear R experts,

I wrote already a couple of weeks ago but did not get any reply.  
Therefore I'll rephrase my question and hope some of you can give me  
some answers or some advice where better to look for an answer.

In general I'm looking for information on how to program functions  
using formulas in order to write a nonparametric multivariate  
regression function with possible clustering of the data.

I would assume that the formulas for my function would have the form
y1 + y2 + y3 ~ x | z
or
Y ~ x | z,

where y1, y2 and y3 are continuous response variables, or Y the matrix  
of responses and x an explaining variable and z a factor indicating  
the cluster membership.

I'm looking for general information about writing functions using  
formulas. But more concrete I would need to know what would be the  
best way to make sure that the response should be multivariate and  
numeric? What is the best way to define the hierarchical structure, is  
it the "|" or is there another, better way? How can I restrict the  
number of hierarchical levels?
And how to use in this context, the data, subset, weights and na.action
arguments?

Thanks in advance,

Klaus

-- 
Klaus Nordhausen
Researcher
Tampere School of Public Health
FIN-33014 University of Tampere

phone:	+358 3 3551 4153
fax:	+358 3 3551 4150
e-mail:	Klaus.Nordhausen at uta.fi

Have a look at the source code for the Formula package.

On Tue, Jul 28, 2009 at 12:58 PM, Klaus
Nordhausen<Klaus.Nordhausen at uta.fi> wrote:
> Dear R experts,
>
> I wrote already a couple of weeks ago but did not get any reply. Therefore
> I'll rephrase my question and hope some of you can give me some answers
or
> some advice where better to look for an answer.
>
> In general I'm looking for information on how to program functions
using
> formulas in order to write a nonparametric multivariate regression function
> with possible clustering of the data.
>
> I would assume that the formulas for my function would have the form
> y1 + y2 + y3 ~ x | z
> or
> Y ~ x | z,
>
> where y1, y2 and y3 are continuous response variables, or Y the matrix of
> responses and x an explaining variable and z a factor indicating the
cluster
> membership.
>
> I'm looking for general information about writing functions using
formulas.
> But more concrete I would need to know what would be the best way to make
> sure that the response should be multivariate and numeric? What is the best
> way to define the hierarchical structure, is it the "|" or is
there another,
> better way? How can I restrict the number of hierarchical levels?
> And how to use in this context, the data, subset, weights and na.action
> arguments?
>
> Thanks in advance,
>
> Klaus
>
> --
> Klaus Nordhausen
> Researcher
> Tampere School of Public Health
> FIN-33014 University of Tampere
>
> phone: ?+358 3 3551 4153
> fax: ? ?+358 3 3551 4150
> e-mail: Klaus.Nordhausen at uta.fi
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

Klaus:
> I wrote already a couple of weeks ago but did not get any reply. Therefore 
> I'll rephrase my question and hope some of you can give me some answers
or
> some advice where better to look for an answer.
>
> In general I'm looking for information on how to program functions
using
> formulas in order to write a nonparametric multivariate regression function
> with possible clustering of the data.
>
> I would assume that the formulas for my function would have the form
> y1 + y2 + y3 ~ x | z
> or
> Y ~ x | z,
>
> where y1, y2 and y3 are continuous response variables, or Y the matrix of 
> responses and x an explaining variable and z a factor indicating the
cluster
> membership.
>
> I'm looking for general information about writing functions using
formulas.
> But more concrete I would need to know what would be the best way to make 
> sure that the response should be multivariate and numeric? What is the best
> way to define the hierarchical structure, is it the "|" or is
there another,
> better way? How can I restrict the number of hierarchical levels?
> And how to use in this context, the data, subset, weights and na.action
> arguments?
The CRAN package "Formula" has some functionality which might assist
you
in what you want to do. It provides several tools for extracting 
information from formulas like
   Y ~ x1 + x2 | z

The version on R-Forge also has some tools for
   y1 + y2 + y3 ~ x1 + x2 | z1 | z2 + z3
etc. But the version on R-Forge is not yet quite stable, however, we hope 
to get something new to CRAN over the summer.

The package just assists you with accessing the different parts of the 
formula. All the checking of types, dimensions, etc. you have to do 
yourself. I hope it solves some of your problems, though...

hth,
Z
> Thanks in advance,
>
> Klaus
>
> -- 
> Klaus Nordhausen
> Researcher
> Tampere School of Public Health
> FIN-33014 University of Tampere
>
> phone:	+358 3 3551 4153
> fax:	+358 3 3551 4150
> e-mail:	Klaus.Nordhausen at uta.fi
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>