[R] step function stops with "subscript out of bounds"

[David A Vavra]

Thanks.  

I guess I still don't understand what's going on. It's not at all intuitive
that the table used should be in the search path. Why is it searching for
the table? Isn't the table already stored in the model? If the documentation
says this, I haven't found it. Needing to know what names I can't use is a
bit disconcerting. I can think of names with high probability of uniqueness
but how could I ever be sure?

Table "X" is built on the fly in a script. What I supplied here was a
portion of the remaining script. I suppose the solution is to store it in an
environment in the search path. 

DAV
	

-----Original Message-----
From: Prof Brian Ripley [mailto:ripley at stats.ox.ac.uk] 
Sent: Thursday, May 17, 2012 6:46 AM
To: David A Vavra
Cc: r-help at r-project.org
Subject: Re: [R] step function stops with "subscript out of bounds"

On 17/05/2012 09:25, David A Vavra wrote:
> I've been having a problem using the step function to evaluate models.
I've
> simplified the code and get the same problem using the dataset Titanic.
The
> relevant code and output is below. The problem disappears (i.e., 'step'
runs
> correctly) if I rerun the code but change the 'loglm' call to explicitly
> reference Titanic instead of X (as in: loglm(as.formula(Y),data=Titanic)).
>
> What is causing this?

A lack of understanding of 'non-standard evaluation'.  X (or at least, 
the X you want) is not visible from the standard search path.

>
> TIA,
> DAV	
>
> ----------
>
>> catn<-function(...) cat(...,"\n")
>> local({  X<-Titanic; print(class(X));
>             Y<-paste('~',paste(names(dimnames(X)),collapse="*"));
>             print(Y);
>             sm<-loglm(as.formula(Y),data=X);
>             catn("SM"); print(sm); catn('running');
>             step(sm,direction='backward')  })

Which will tell you
Error in eval(expr, envir, enclos) : could not find function "loglm"

If you correct that and use a vanilla session you will get

Error in loglm(formula = ~Class + Sex + Age + Survived + Class:Sex + 
Class:Age +  :
   object 'X' not found

which is more informative.

So the solution is to

- use less easily masked names than 'X'.
- ensure the data object is visible on the search path.


> Output:
>
> [1] "table"
> [1] "~ Class*Sex*Age*Survived"
> SM
> Call:
> loglm(formula = as.formula(Y), data = X)
>
> Statistics:
>                   X^2 df P(>  X^2)
> Likelihood Ratio   0  0        1
> Pearson          NaN  0        1
> running
> Start:  AIC=64
> ~Class * Sex * Age * Survived
>
> Error in loglin(data, margins, start = start, fit = fitted, param = param,
> :
>    subscript out of bounds
>
> Enter a frame number, or 0 to exit
>
> 1: local({
>      X<- Titanic
>      print(class(X))
>      Y<- paste("~", paste(names(dimnames(X)), collapse = "*"))
>      print(Y)
>      sm<- loglm(as.formula(Y), data = X
>   2: eval.parent(substitute(eval(quote(expr), envir)))
>   3: eval(expr, p)
>   4: eval(expr, envir, enclos)
>   5: eval(quote({
>      X<- Titanic
>      print(class(X))
>      Y<- paste("~", paste(names(dimnames(X)), collapse = "*"))
>      print(Y)
>      sm<- loglm(as.formula(Y), dat
>   6: eval(expr, envir, enclos)
>   7: #1: step(sm, direction = "backward")
>   8: #1: drop1(fit, scope$drop, scale = scale, trace = trace, k = k, ...)
>   9: #1: drop1.default(fit, scope$drop, scale = scale, trace = trace, k =
k,
> ...)
> 10: #1: update(object, as.formula(paste("~ . -", tt)), evaluate = FALSE)
> 11: #1: update.loglm(object, as.formula(paste("~ . -", tt)), evaluate =
> FALSE)
> 12: #1: eval.parent(call)
> 13: #1: eval(expr, p)
> 14: #1: eval(expr, envir, enclos)
> 15: #1: loglm(formula = ~Class + Sex + Age + Survived + Class:Sex +
> Class:Age + Sex:Age + Class:Survived + Sex:Survived + Age:Survived +
> Class:Sex:Age + Class:
> 16: #1: loglm1(formula, data, ..., .call = .call, .formula = .formula)
> 17: #1: loglm1.default(formula, data, ..., .call = .call, .formula =
> .formula)
> 18: #1: loglin(data, margins, start = start, fit = fitted, param = param,
> eps = eps, iter = iter, print = print)
>
> Selection: 0
>
> ______________________________________________
> 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.


-- 
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