Problem with logLik.lm()

["pierre duchesne" <pierre.duchesne@hec.ca>]

Dear S+ users,

Under 
S-PLUS 2000 Professional Edition Release 2 for Microsoft Windows : 1999,
with Windows 95, I wrote the following function
to compare two models according BIC criteria.

myfun _ function(){
 set.seed(1)
 x _ arima.sim(n=100, model=list(ar=.9))
 model.1 _ lm( x[3:100] ~ x[2:99]  )
 model.2 _ lm( x[3:100] ~ x[2:99] + x[1:98] )
 BIC(model.1,model.2)
}

When I run the function, I have the following error
message:

> myfun()
myfun()
Error in function(object, ...): Object "x" not found
>

So I looked in the code of the function logLik.lm(),
and I found that there is an update() operation.
# ==== part of the code of lokLik.lm() ==============
        if(is.null(object$qr))
                object <- update(object, qr = T)
# ===================================================

So if I modify my original code in the following way:

myfun.qr _ function(){
 set.seed(1)
 x _ arima.sim(n=100, model=list(ar=.9))
 model.1 _ lm( x[3:100] ~ x[2:99], qr=T )
 model.2 _ lm( x[3:100] ~ x[2:99] + x[1:98], qr=T )
 BIC(model.1,model.2)
}

all is now fine.  But I avoid the error message since there is no
update step!  It seems that in the update step, the function
is not able to find the data.  Is it related to a problem
of data.frame?  Is there a way to tell to logLik.lm()
where are the data in my original function myfun()?

I tried the following but that does not work:
myfun2 _ function(){
 set.seed(1)
 x _ arima.sim(n=100, model=list(ar=.9))
 mydata _ cbind(x[3:100],x[2:99],x[1:98])
 mydata _ as.data.frame(mydata)
 model.1 _ lm( x[3:100] ~ x[2:99], data=mydata  )
 model.2 _ lm( x[3:100] ~ x[2:99] + x[1:98] , data=mydata)
 BIC(model.1,model.2)
}

I presume that I do not understand well the data.frame in that
situation.

Thanks for your help,

P.S. The drastic function myfun3() works, but I prefer to avoid
     the 'terrible' operator "<<-".
myfun3 _ function(){
 set.seed(1)
 x <<- arima.sim(n=100, model=list(ar=.9))
 model.1 _ lm( x[3:100] ~ x[2:99]  )
 model.2 _ lm( x[3:100] ~ x[2:99] + x[1:98] )
 BIC(model.1,model.2)
}

-- 
Au plaisir, Pierre. 
_____________________________________________________________________ 
Pierre Duchesne, professeur adjoint, 
Service de l'enseignement des methodes quantitatives de gestion (MQG) 
Ecole des Hautes Etudes Commerciales          (http://www.hec.ca/mqg) 
3000, chemin de la Cote-Sainte-Catherine, Montreal(QC) Canada H3T 2A7 
Bureau: 4.821            Tel:(514)340-6503          Fax:(514)340-5634 
Page WEB: http://www.hec.ca/profs/pierre.duchesne.html 
=====================================================================