R devel - Jan 2007 - How to evaluate an lm() object for generating warning statement in a function

Hi, there

I tried this post on R-help but did not generate any replies, so I thought 
I might try the waters here since it's a more programming-based group.

I have written a function that will allow me to calculate the variance
components for a model II (random effects) single factor ANOVA (perhaps 
this is
me re-inventing the wheel, but I handn't yet come across anything in R that
does this for me, but I'm sure people will let me know from this posting if
there are). The function takes as it's argument an object which holds the
results of an lm call. e.g.,

object1<-lm(y~x)

The function is as follows (some comments included):

comp.var.estimates<-function(object.lm)
                 {
                 anovmod<-anova(object.lm) #get the anova table for the lm
                 MStreat<-anovmod[1,3]; MSErr<-anovmod[2,3] #extract Mean 
Squares
                 dataframe<- as.data.frame(object.lm[13]) #gets the data 
that went into the lm
                 ni <- tapply(dataframe[,1], dataframe[,2], length) #number 
of cases per treatment level
                 nisq<-ni^2
                 no<-(1/(length(ni)-1))*(sum(ni)-(sum(nisq)/sum(ni))) 
#required to calculate variance components
                 s2a<-((MStreat-MSErr)/no)
                 stot<-s2a + MSErr
                 treatvar<-s2a/stot*100  #calculate variance components as 
a percentage of the total
                 errorvar<-MSErr/stot*100
                 list(treat.var.comp=s2a, 
err.var.comp=MSErr,
                 p.var.treat=treatvar, p.var.err=errorvar)
                 }
comp.var.estimates(object1)

I'd like to include a "warning" statement in the function that
returns something like 'function requires arguments that are objects of the 
form obj<-lm
(y~x)', but I need a case to evaluate the object against in order to throw
the
warning.

Any advice?

I feel like my best opportunity is after the first line into the function,
where I ask for the ANOVA table. Since this is a 2 X 5 table, presumably I
should be able to evaluate it against the size of that table? Any thoughts on
how to check that? I welcome any suggestions.

Cheers,

Mike




Michael Rennie
Ph.D. Candidate, University of Toronto at Mississauga
3359 Mississauga Rd. N.
Mississauga, ON  L5L 1C6
Ph: 905-828-5452  Fax: 905-828-3792
www.utm.utoronto.ca/~w3rennie 
	[[alternative HTML version deleted]]

There are two options, you can check to see if 'object.lm' inherits from
"lm"
using 'inherits()' or make comp.var.estimates a generic and then make a
method
for "lm" objects.

-roger

Michael Rennie wrote:
> Hi, there
> 
> I tried this post on R-help but did not generate any replies, so I thought 
> I might try the waters here since it's a more programming-based group.
> 
> I have written a function that will allow me to calculate the variance
> components for a model II (random effects) single factor ANOVA (perhaps 
> this is
> me re-inventing the wheel, but I handn't yet come across anything in R
that
> does this for me, but I'm sure people will let me know from this
posting if
> there are). The function takes as it's argument an object which holds
the
> results of an lm call. e.g.,
> 
> object1<-lm(y~x)
> 
> The function is as follows (some comments included):
> 
> comp.var.estimates<-function(object.lm)
>                  {
>                  anovmod<-anova(object.lm) #get the anova table for the
lm
>                  MStreat<-anovmod[1,3]; MSErr<-anovmod[2,3] #extract
Mean
> Squares
>                  dataframe<- as.data.frame(object.lm[13]) #gets the data
> that went into the lm
>                  ni <- tapply(dataframe[,1], dataframe[,2], length)
#number
> of cases per treatment level
>                  nisq<-ni^2
>                  no<-(1/(length(ni)-1))*(sum(ni)-(sum(nisq)/sum(ni))) 
> #required to calculate variance components
>                  s2a<-((MStreat-MSErr)/no)
>                  stot<-s2a + MSErr
>                  treatvar<-s2a/stot*100  #calculate variance components
as
> a percentage of the total
>                  errorvar<-MSErr/stot*100
>                  list(treat.var.comp=s2a, 
> err.var.comp=MSErr,
>                  p.var.treat=treatvar, p.var.err=errorvar)
>                  }
> comp.var.estimates(object1)
> 
> I'd like to include a "warning" statement in the function
that
> returns something like 'function requires arguments that are objects of
the
> form obj<-lm
> (y~x)', but I need a case to evaluate the object against in order to
throw the
> warning.
> 
> Any advice?
> 
> I feel like my best opportunity is after the first line into the function,
> where I ask for the ANOVA table. Since this is a 2 X 5 table, presumably I
> should be able to evaluate it against the size of that table? Any thoughts
on
> how to check that? I welcome any suggestions.
> 
> Cheers,
> 
> Mike
> 
> 
> 
> 
> Michael Rennie
> Ph.D. Candidate, University of Toronto at Mississauga
> 3359 Mississauga Rd. N.
> Mississauga, ON  L5L 1C6
> Ph: 905-828-5452  Fax: 905-828-3792
> www.utm.utoronto.ca/~w3rennie 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
> 
-- 
Roger D. Peng  |  http://www.biostat.jhsph.edu/~rpeng/