%
% NOTE -- ONLY EDIT THE .Rnw FILE!!!  The .tex file is
% likely to be overwritten.
%
\documentclass[12pt]{article}

\usepackage{amsmath}
\usepackage[authoryear,round]{natbib}
\usepackage{hyperref}


\textwidth=6.2in
\textheight=8.5in
%\parskip=.3cm
\oddsidemargin=.1in
\evensidemargin=.1in
\headheight=-.3in

\newcommand{\scscst}{\scriptscriptstyle}
\newcommand{\scst}{\scriptstyle}


\newcommand{\Rfunction}[1]{{\texttt{#1}}}
\newcommand{\Robject}[1]{{\texttt{#1}}}
\newcommand{\Rpackage}[1]{{\textit{#1}}}
\newcommand{\Rmethod}[1]{{\texttt{#1}}}
\newcommand{\Rfunarg}[1]{{\texttt{#1}}}
\newcommand{\Rclass}[1]{{\textit{#1}}}

\textwidth=6.2in

\bibliographystyle{plainnat} 
 
\begin{document}
%\setkeys{Gin}{width=0.55\textwidth}

\title{Notes on non-syntactic names}
\author{VJ Carey}
\maketitle

\section{Objective}

We want to be able to handle non-syntactic names in formulae
with minimal added infrastructure.  Let's see an example.
<<getALL>>=
if (!("ALL" %in% search())) library(ALL)
if (!exists("ALL")) data(ALL)
lit = t(exprs(ALL)[1:5,])
lit[1:3,1:4]
dlit = data.frame(lit, neg=factor(ALL$mol.biol=="NEG"))
dlit[1:3,1:4]
dlit2 = data.frame(lit, neg=factor(ALL$mol.biol=="NEG"), check.names=FALSE)
dlit2[1:3,1:4]
@

\section{GLM and others that work without change}
<<dog>>=
library(randomForest)
library(MASS)
glm(neg~., data=dlit, fam=binomial)
glm(neg~., data=dlit2, fam=binomial) 
randomForest(neg~., data=dlit2, fam=binomial) 
lda(neg~., data=dlit2, fam=binomial) 
@

\section{RPART}
<<dor>>=
library(rpart)
rpart(neg~., data=dlit)
lk = try(rpart(neg~., data=dlit2))
lk
@
In rpart, we have

\begin{verbatim}
 lapply(m[attr(Terms, "term.labels")], tfun)
\end{verbatim}
and the backticks need to be removed for the selection in m.  In other words
defining
\begin{verbatim}
 cleanTick = function(x) gsub("`", "", x)
\end{verbatim}
and using
\begin{verbatim}
 lapply(m[cleanTick(attr(Terms, "term.labels"))], tfun)
\end{verbatim}
moves us along a bit.  We also need a change to rpart.matrix, at
\begin{verbatim}
    else x <- as.matrix(frame[cleanTick(predictors)])
\end{verbatim}

\section{NNET}
<<dnn>>=
library(nnet)
nnet(neg~., data=dlit, size=2)
lk = try(nnet(neg~., size=2, data=dlit2))
lk
@
if we modify
\begin{verbatim}
> .getXlevels
function (Terms, m) 
{
    xvars <- sapply(attr(Terms, "variables"), deparse, width.cutoff = 500)[-1]
    if ((yvar <- attr(Terms, "response")) > 0) 
        xvars <- cleanTick(xvars[-yvar]) # change here
    if (length(xvars) > 0) {
        xlev <- lapply(m[xvars], levels)
        xlev[!sapply(xlev, is.null)]
    }
    else NULL
}
\end{verbatim}
the function works.

\section{Standalone approach}

<<sta>>=
set.seed(123)
x = matrix(rnorm(40),nr=8)
colnames(x) = paste(1001:1005,"_at", sep="")
X = data.frame(x, check.names=FALSE)
glm(`1005_at`~., data=X)
library(MASS)
lda(I(`1005_at`>0)~., data=X)
library(nnet)
try(nnet(I(`1005_at`>0)~., data=X, size=2))
library(rpart)
try(rpart(I(`1005_at`>0)~., data=X))
@

\begin{verbatim}

traceback for nnet:

Error in `[.data.frame`(m, xvars) : undefined columns selected

Enter a frame number, or 0 to exit   

1: try(nnet(I(`1005_at` > 0) ~ ., data = X, size = 2))
2: nnet(I(`1005_at` > 0) ~ ., data = X, size = 2)
3: nnet.formula(I(`1005_at` > 0) ~ ., data = X, size = 2)
4: .getXlevels(Terms, m)
5: lapply(m[xvars], levels)
6: is.vector(X)
7: m[xvars]
8: `[.data.frame`(m, xvars)

traceback for rpart

Error in `[.data.frame`(frame, predictors) : 
        undefined columns selected

Enter a frame number, or 0 to exit   

1: try(rpart(I(`1005_at` > 0) ~ ., data = X))
2: rpart(I(`1005_at` > 0) ~ ., data = X)
3: rpart.matrix(m)
4: as.matrix(frame[predictors])
5: frame[predictors]
6: `[.data.frame`(frame, predictors)

\end{verbatim}

\begin{verbatim}
solution for nnet:

> .getXlevels2
function (Terms, m) 
{
    detick = function(x) gsub("`","",x)
    xvars <- sapply(attr(Terms, "variables"), deparse, width.cutoff = 500)[-1]
    if ((yvar <- attr(Terms, "response")) > 0) 
        xvars <- detick(xvars[-yvar])
    if (length(xvars) > 0) {
        xlev <- lapply(m[xvars], levels)
        xlev[!sapply(xlev, is.null)]
    }
    else NULL
}
\end{verbatim}



\end{document}