s-news
[Top] [All Lists]

Re: Burned by factors

To: <Bill.Venables@csiro.au>, <bates@stat.wisc.edu>, <kw.statr@gmail.com>
Subject: Re: Burned by factors
From: "Alan Hochberg" <alan.hochberg@prosanos.com>
Date: Mon, 17 Mar 2008 09:21:15 -0400
Cc: <s-news@lists.biostat.wustl.edu>
In-reply-to: <B998A44C8986644EA8029CFE6396A9240100CE1D@exqld2-bne.nexus.csiro.au>
References: <c968588d0803051012x4be861f7pbf489ae363b0983e@mail.gmail.com> <40e66e0b0803151310r3fb5c03fq2e50cb7892c60853@mail.gmail.com> <B998A44C8986644EA8029CFE6396A9240100CE1D@exqld2-bne.nexus.csiro.au>
Thread-index: AciG2NM7uGJgFpT3T9iWTG/iAEFFZwAJBJbQAEzrIYA=
This is clearly an emotional issue for those of us who have experienced
"correct" but unexpected behavior from the interaction of factors with
character and numeric data types.  In my experience, this usually happens
when a deadline is near on a critical project.

I'm wondering aloud if it would be feasible for R/S compilers and run-time
systems to have a mode where they issue warnings in situations where
confusion is likely: For instance, when all the levels for a factor are
valid text representations of numbers, or when a factor in its numeric guise
is compared to a character string.  Would anyone from the R/S infrastructure
world care to comment on the feasibility/desirability of this?

Thanks,

Alan

Alan Hochberg
VP, Research
ProSanos Corporation
225 Market St. Ste. 502, 
Harrisburg, PA 17101
Tel 717-635-2124 * Fax 717-635-2575


<Prev in Thread] Current Thread [Next in Thread>