%<*paper|techreport|present>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% TITLE:   On Fragile Grounds:
%%          A replication of "Are Muslim immigrants
%%          different in terms of cultural integration?"
%% AUTHORS: Mahmood Arai, Jonas Karlsson and Michael
%%          Lundholm
%% CONTACT: mahmood.arai@ne.su.se
%%          jonas.karlsson@sofi.su.se
%%          michael.lundholm@ne.su.se
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% This batch file can be used to generate the relevant
%% files performing the estimations, presentations and
%% documentations of the project by using the R package
%% Sweave, the LaTeX DOCSTRIP utility and the standard LaTeX
%% family of software using the UN*X commands
%%
%% echo 'Sweave("araietal_source.Rnw")' | R --vanilla --quiet
%% echo 'source("script.R")' | R --vanilla --quiet
%%
%% or from within R using
%%
%% Sweave("araietal_source.Rnw")
%% source("script.R")
%%
%% The file araietal_source.Rnw contains all source code
%% and run through Sweave it generates the files
%% araietal_source.tex, script.R, araietal_source.R,
%% araietal.ins and araietal.bib. When the file script.R is
%% run first LaTeX's DOCSTRIP utility is run via
%% araietal.ins and the files araietal_paper.tex,
%% araietal_techreport.tex and araietal_present.tex are
%% generated. The script then continues to process these
%% files with the LaTeX family of programs, using the
%% BibTex bibliography database file araietal.bib, to
%% generate the corresponding PDF-files for paper,
%% technical documentation and presentation.
%%
%% Accordingly the file contains three levels of markup
%% 1. noweb markup to define the code chunks evaluated
%%    when "araietal_source.Rnw" is run through R via Sweave.
%% 2. DOCSTRIP markup to define the conditional LaTeX code
%%    to be shipped to the different generated LaTeX files.
%%    Conditioning is made on the tags "paper" (which
%%    generates araietal_paper.tex and contains our
%%    research result in an article, "techreport" (which
%%    generates araietal_techreport.tex and contains a
%%    complete technical documentation of our research) and
%%    "present" (which generates araietal_present.tex and
%%    contains a slide presentation of our research).
%%    when "latex araietal.ins" is run.
%% 3. Standard LaTeX markup to be considered when the
%%    generated LaTeX files are run with latex/pdflatex.
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%</paper|techreport|present>
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Setting document classes for LaTeX conditional on the
%% type of document (paper, techreport or present)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%<*paper>
\documentclass[a4paper,11pt]{article}
%</paper>
%<*techreport>
\documentclass[a4paper,11pt]{article}
%</techreport>
%<*present>
\documentclass[style=horatio,mode=present,%
     paper=screen]{powerdot}
\pdsetup{method=direct}
%</present>
%<*paper|techreport|present>
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Loading LaTeX packages
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
\usepackage[latin1]{inputenc}
\usepackage[T1]{fontenc}
\usepackage[round]{natbib}
\usepackage{Sweave,fancyvrb,color,url,hyperref,multirow,%
     paralist,rotating,}
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% LaTeX definitions
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\definecolor{Red}{rgb}{0.5,0,0}
\definecolor{Blue}{rgb}{0,0,0.5}
\hypersetup{%
    breaklinks = {true},
    colorlinks = {true},
    linkcolor = {Blue},
    citecolor = {Blue},
    urlcolor = {Red}
    }
\newcommand{\code}[1]{{\upshape\mdseries\ttfamily #1}}
\newcommand{\proglang}[1]{{\upshape\mdseries\sffamily #1}}
\newcommand{\pkg}[1]{{\upshape\bfseries\rmfamily #1}}
\newcommand{\email}[1]{\href{mailto:#1}%
     {\normalfont\texttt{#1}}}
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Setting options for R/Sweave
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\setkeys{Gin}{width=\textwidth}
\SweaveOpts{keep.source=TRUE, echo=TRUE}
<<Options,echo=FALSE,quite=TRUE>>=
options(width=90,scipen=3,digits=4)
@
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Defining conditional title pages
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%</paper|techreport|present>
%<*paper>
\title{\LARGE\bfseries On Fragile Grounds:\\
       A replication of \emph{Are Muslim immigrants
       different in terms of cultural integration?}}
%</paper>
%<*techreport>
\title{\LARGE\bfseries On Fragile Grounds:\\
       A replication of \emph{Are Muslim immigrants
       different in terms of cultural integration?
       }\\Technical documentation}
%</techreport>
%<*present>
\title{On Fragile Grounds:\\
       A replication of "Are Muslim immigrants
       different in terms of cultural integration?"}
%</present>
%<*paper|techreport>
\author{Mahmood Arai,\footnote{Corresponding author.
    Department of Economics and SULCIS, Stockholm
    University, SE 106~91 Stockholm, Sweden,
    \email{mahmood.arai@ne.su.se}.} { }Jonas
    Karlsson\footnote{The Institute for Social Research and
    SULCIS, Stockholm University,
    \email{jonas.karlsson@sofi.su.se}.}  { }and Michael
    Lundholm\footnote{Department of Economics, Stockholm
    University, \email{michael.lundholm@ne.su.se}.}}
%</paper|techreport>
%<*present>
\author{Mahmood Arai, Jonas Karlsson and Michael Lundholm}
%</present>
%<*paper|techreport|present>
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Starting the LaTeX document and creating title pages
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{document}
%</paper|techreport|present>
%<*present>
\sffamily
%</present>
%<*paper|techreport|present>
\maketitle
%</paper|techreport|present>
%<*paper>
\begin{abstract}
This study is a replication of  ``Are Muslim Immigrants
Different in terms of Cultural Integration?'' by Alberto
Bisin, Eleonora Patacchini, Thierry Verdier and Yves Zenou,
published in Journal of European Economic Association, 6,
445-456, 2008.


\citet{Bisin08} report that they have 5963 observations in
their study.
Using their empirical setup, we can only identify
\input{crossreference3.tex}
relevant observations in the original data. After removing
missing values we are left with
\input{crossreference4.tex}
observations. We cannot replicate any of their results and
our estimations yield no support for their claims.


\end{abstract}
%</paper>
%<*techreport>
\begin{abstract}
This is a technical documentation of \citet{Araietal08a}
which replicates ``Are Muslim Immigrants Different in terms
of Cultural Integration?'' by Alberto Bisin, Eleonora
Patacchini, Thierry Verdier and Yves Zenou, published in
Journal of European Economic Association, 6, 445-456, 2008.



\citet{Bisin08} report that they have 5963 observations in
their study. Using their empirical setup,
we can only identify
\input{crossreference3.tex}
relevant observations in the original data. After removing
missing values we are left with
\input{crossreference4.tex}
observations. We cannot replicate any of their results and
our estimations yield no support for their
claims.


\end{abstract}
\newpage
\tableofcontents
\newpage
%</techreport>
%<*paper|techreport|present>
%%
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Document content starting
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%</paper|techreport|present>


%<*present>
\begin{slide}{Introduction}
\vspace{\stretch{1}}
\begin{itemize}
\item This study is a replication of  ``Are Muslim
Immigrants Different in terms of Cultural Integration?'' by
Alberto Bisin, Eleonora Patacchini, Thierry Verdier and Yves
Zenou, published in Journal of European Economic
Association, 6, 445-456, 2008.
\end{itemize}
\vspace{\stretch{1}}
\end{slide}

%</present>



%<*paper|techreport>
\section{Introduction}

This is a replication of the empirical
results reported in \citet{Bisin08}.  They use British
data and analyse how Muslims and non-Muslims
differ in cultural integration measured as (i)
\emph{Importance of Religion}, (ii) \emph{Attitude
Towards Inter Marriage} and (iii) \emph{Importance of
Racial Composition in Schools}.\footnote{To facilitate
comparability we use the same labels on the variables as
\citet{Bisin08}.}
%</paper|techreport>

%<*present>
\begin{slide}{Their claims}
\vspace{\stretch{1}}
In the abstract of their paper they write:
\begin{center}
\begin{quote}
``\dots Muslims integrate less and more slowly than
non-Muslims. \dots We also find no evidence that
segregated neighbourhoods breed intense religious and
cultural identities for ethnic minorities, especially for
Muslims.''

``\dots On the contrary, \dots
intense identities in our data are more prominent
in relatively mixed neighbourhoods.''
(p. 446)
\end{quote}
\end{center}
\vspace{\stretch{1}}
\end{slide}
%</present>

%<*paper|techreport>
In the abstract of their paper they write:
\begin{quote}
``\dots Muslims integrate less and more slowly than
non-Muslims. \dots We also find no evidence that
segregated neighbourhoods breed intense religious and
cultural identities for ethnic minorities, especially for
Muslims.'' \citep[p. 245]{Bisin08}
 \end{quote}

We wanted to check the robustness
of their results when considering the ethnic and religious
heterogeneity within both groups, Muslims and non-Muslims.
Among other things, we were concerned about the measures of
cultural values used in the paper. These measures capture
ethnic and  religious attributes in different degrees for
different groups. For example the variable \emph{Attitudes
towards Inter-Marriage} with the majority UK population
captures only inter-ethnic marriage for the Christian ethnic
minorities but both inter-ethnic and inter-religious
marriage for Muslims.
%</paper|techreport>

%<*present>
\begin{slide}[toc=]{Our concerns}
\vspace{\stretch{1}}
\begin{enumerate}\setcounter{enumi}{1}
\item The measures of cultural values capture ethnic and
religious attributes in different degrees for different
groups.
\end{enumerate}
\vspace{\stretch{1}}
\end{slide}
%</present>

%<*paper|techreport>
However, already an initial inspection of data disclosed
that the number of observations in \citet{Bisin08}
exceeded the total number of observations in the ethnic
minority sample. We communicated this
to the authors and they answered that there were
some coding errors. We have received revised codes and a
revised version of their specifications and tables. Their
revised codes yield fewer observations than the sample in
the published version, but still more than we can identify
in the relevant sample of the original data. As far as we
can see, a source of the large number of observations
in their revised codes is that dummy variable definitions
include  observations with missing values in the reference
categories (defined as zeros). The underlying codes to the
published paper were, however, not made available and the
exact nature of the original errors are therefore unknown
to us.

\citet{Bisin08} report that they have 5963 observations in
their study, whereas the ethnic minority sample in
\citet{Berthoud} consists of \input{crossreference1.tex}
observations. Implementing their empirical setup, we can
only identify \input{crossreference3.tex} relevant
observations in the original data. After removing missing
values we are left with \input{crossreference4.tex}
observations. Using the remaining sample and running their
specifications, we find no results that support their
claims. Our replication therefore stopped here and we did
not perform any sensitivity analysis. The great loss of
observations implies that the remaining sample is most
likely not representative. Therefore, we hesitate to draw
inference from the regressions results.
%</paper|techreport>

%<*present>
\begin{slide}{Our results}
\vspace{\stretch{1}}

\begin{itemize}
\item In their paper the number of observations
is 5963 which is $\input{crossreference0.tex}$ percent of
the number of observations in the relevant sample of
the original data ($\input{crossreference3.tex}$).
\item Implementing their variable definitions we loose
$\input{crossreference2.tex}$ percent of the original
sample and have only $\input{crossreference4.tex}$
observations.
\item Using the remaining sample and running their
specifications, we find no results that support their
claims.
\end{itemize}
\vspace{\stretch{1}}
\end{slide}

%</present>

%<*paper|techreport>

In this paper we only document the replication and report
and comment results using the variable definitions, the
variable names and the specifications used in
\citet{Bisin08}. We choose a procedure that makes it easy to
reproduce our results. Influenced by \cite{Koenker07}, we
use an integrated approach where data management,
estimations, and the text that rely on these computations
are all integrated in one single file. This strategy has the
advantage that it makes is easy to adjust the codes and
automatically generate a revised version of the paper.
%</paper|techreport>





%<*paper>
For details of our analysis of the data and implementation
 of the variable definitions, see \citet{Araietal08b},
 which is a technical companion to this paper.
%</paper>
%<*paper|techreport>
All data analysis is made in \proglang{R} \citep{Rcore}
and all code files related to this project can be found
on \url{http://people.su.se/~lundh/fragile_grounds/}.
%</paper|techreport>




%<*techreport>

In this technical documentation we present
our results in greater detail, but also all our working
procedures, variables definitions etc. In addition the
central part of our codes are included with typeset
comments. This is done as an attempt to implement Literate
Statistical Programming.

The remainder of the paper is organised as follows.
The data and variable definitions are described in Section
 \ref{Data}. Regression results are presented in Section
 \ref{Regression}. The paper is concluded in Section
 \ref{Discussion}. Finally the production procedure is
described in Section \ref{sec:prodnotes}.
%</techreport>


%<*paper>

The remainder of the paper is organised as follows.
The data are described in Section \ref{Data}. Regression
results are presented in Section \ref{Regression} and
finally the paper is concluded in Section \ref{Discussion}.
%</paper>


%<*present>
\begin{slide}[toc=What we do]{What we wanted to do, what we
did and what we didn't}
\vspace{\stretch{1}}

\begin{itemize}
\item We wanted to:
     \begin{itemize}
     \item replicate their study,
     \item check the robustness of their results
     \end{itemize}
\item What we do
     \begin{itemize}
     \item document the replication and report
      and comment results using their variable definitions
      etc
      \end{itemize}
\item What we didn't do:
      \begin{itemize}
      \item almost no sensitivity checking
      \end{itemize}
\end{itemize}
\vspace{\stretch{1}}
\end{slide}


%</present>


%<*paper|techreport>
\section{Data and variable description}
\label{Data}
%</paper|techreport>

%<*present>
\begin{slide}{Data}
\vspace{\stretch{1}}
\begin{itemize}
\item Fourth National Survey of Ethnic Minorities 1993-1994
\item UK Data Archive (UKDA) via Athens.
\end{itemize}
\vspace{\stretch{1}}
\end{slide}
%</present>





%<*techreport>
\subsection{Data}
%</techreport>
%<*paper|techreport>
The data set is the \emph{Fourth National Survey of
Ethnic Minorities 1993-1994} (FNSEM); see
\citep{Berthoud}.\footnote{The data can be accessed
from the UK Data Archive (UKDA) via Athens. UK Data
Archive is found at \url{http://www.data-archive.ac.uk/}
and Athens at \url{http://www.athens.ac.uk/}.}
%</paper|techreport>
%<*techreport>
It's main objective were:
\begin{quote}
\begin{itemize}
\item ``to describe the social and economic conditions
of Britain's main ethnic minority groups, including
their health, and to compare these with the social
and economic conditions of the white majority
\item to assess changes over time through comparisons
with other work
\item to show how the position of ethnic minority
groups is related to the social and ethnic compositions
of the areas in which they live
\item to explore diversity among different ethnic
minority groups
\item to describe perceptions and experience of
racial discrimination and social harassment''
\end{itemize}
\citet{Berthoud}
\end{quote}

For our coding we have used \citet{FNSEM}, which contains
the project instructions, and \citet{FNSEM93b}, which
is the data description file included in the  files
obtained when the entire data set is downloaded
from the UKDA.

In the following we present how the original data are used
to define the data set used in the estimations. We present
extracts of our \proglang{R} code \citep{Rcore} with
extensive comments and discussions. For details about
our working procedures and how we document the
research, see section \ref{sec:prodnotes} on page
\pageref{sec:prodnotes}. In the code chunks ``$>$''
denotes the \proglang{R} prompt and ``$+$''
continuation of the previous line.

\subsection{Reading data and selecting variables}
We load  package \pkg{foreign} to read STATA data format.
Data is read from the unpacked Stata-version of the
data and ``\verb+_+'' in variable names are converted
to ``$.$''.

<<Reading data>>=
library(foreign)
FNSEM <- data.frame(read.dta("3685.dta",
     convert.underscore=TRUE))
@
<<Selecting variables, echo=FALSE>>=
FNSEM<-subset(FNSEM,select=c(
     a1an,a1e,a3,a4n,area,ethnic,
     f14a,f14b,f15a,f15bn:f15dn,f16a,f16b1n:f16b3n,
     hh2a.s,hh2c.b,hh2c.c,hh2c.d,hh2c.e,hh2c.f,
          hh2c.g,hh2c.h,hh2c.i,hh2c.j,hh2c.k,
          hh2c.l,hh2c.m,hh5d.s,hh40,
     j1b,j2,j3occ,j55a,j63a,
     q1,q3,question,
     s6,s7,s9,s12a,s12b,s12c,s24a,s34a,s34b,s39,
     v1a,v1b,v1c,v1d,v9a,
     weightis,wown,wunemp,
     year,
     ..q2a1,..q2a2,..q2a3,..q2a4,..q2a5,..q2a6,
          ..q2a7,..q2a8,..q2a9,..q2a10,..q2a11,
          ..q2a12,..q2a13,..q2a14,..q2a15,
          ..q2a16,..q2a17,..q2a18,..q2a19,
          ..q2a20,..q2a21,..q2a22,..q2a23,
          ..q2a24,..q2a25,..q2a26,..q2a27,
          ..q2a28,..q2a29,..q2a30,..q2a31,
          ..q2a32,..q2a33,
     ..s12f1,..s12f2, ..s12f3, ..s12f4, ..s12f5,
             ..s12f6, ..s12f7, ..s12f8, ..s12f9,
             ..s12f10,..s12f11,..s12f12,..s12f13,
             ..s12f14,..s12f15,..s12f16,..s12f17,
             ..s12f18,
     ..s12g1,..s12g2, ..s12g3, ..s12g4, ..s12g5,
             ..s12g6, ..s12g7, ..s12g8, ..s12g9,
             ..s12g10,..s12g11,..s12g12,..s12g13,
             ..s12g14,..s12g15,..s12g16,..s12g17,
             ..s12g18,
     ..s12h1,..s12h2, ..s12h3, ..s12h4, ..s12h5,
             ..s12h6, ..s12h7, ..s12h8, ..s12h9,
             ..s12h10,..s12h11,..s12h12,..s12h13,
             ..s12h14,..s12h15,..s12h16,..s12h17,
             ..s12h18,
     ..s12i1,..s12i2, ..s12i3, ..s12i4, ..s12i5,
             ..s12i6, ..s12i7, ..s12i8, ..s12i9,
             ..s12i10,..s12i11,..s12i12,..s12i13,
             ..s12i14,..s12i15,..s12i16,..s12i17,
             ..s12i18
))
@
After reading the data we select a subset of variables to
be used. This code is in \code{araietal\_source.Rnw} but not
shown here. It is also available in
\code{araietal\_source.R}.

%</techreport>

%<*present>
\begin{wideslide}{Sample}
\vspace{\stretch{1}}
\footnotesize{
<<Ethnic, echo=TRUE>>=
t(table(FNSEM$ethnic,exclude=c()))
@
}
\begin{itemize}
\item Remove Whites
\end{itemize}
\vspace{\stretch{1}}
\end{wideslide}


\begin{slide}[toc=]{Sample (cont.)}
\vspace{\stretch{1}}
<<Religious, echo=TRUE>>=
table(FNSEM$s6,exclude=c())
@
\begin{itemize}
\item Remove non--religious (s6=2)
\end{itemize}
\vspace{\stretch{1}}
\end{slide}


\begin{slide}[toc=]{Sample (cont.)}
\vspace{\stretch{1}}
<<Singles, echo=TRUE>>=
table(FNSEM$a1e,exclude=c())
@

\begin{itemize}
\item Remove singles (a1e=3)
\end{itemize}
\vspace{\stretch{1}}
\end{slide}

\begin{slide}[toc=]{Sample (cont.)}
\vspace{\stretch{1}}
\begin{itemize}
\item One of three possible questionnaires (green (catageory
1), yellow (category 2)
and pink (category 3)) were used.
\item Questions involved in the study are in the green
questionnaire.
\end{itemize}
<<Questionnaire, echo=TRUE>>=
table(FNSEM$question,exclude=c())
@

\begin{itemize}
\item Remove pink and yellow (question=2,3)
\end{itemize}
\vspace{\stretch{1}}
\end{slide}

%</present>

%<*techreport>
\subsection{Defining the relevant (ethnic minority) sample}
\label{sec:definingsample}
The data consist of two samples, \emph{Ethnic Minorities}
and \emph{Whites}. We are only interested in the former
and remove all Whites. The variable \code{ethnic} indicates
ethnic group of the individual according to the British
standard and is used for this purpose.

One of the three measures of cultural integration in
\citet{Bisin08} is \emph{Importance of Religion}.
Whether a respondent has a religion or belongs to a
church is registered in question \code{s6}. Those who do
not have a religion or do not belong to a church are coded
$2$; we remove these observations from the sample since
they cannot be classified in a religious group.

A variable used in \citet{Bisin08} concerns the role of the
respondent and his or her parents about choosing the
respondent's husband/wife. Since this information is only
available for married and previously married persons, the
unmarried persons are removed from the sample.

Furthermore, respondents were faced with one out of three
questionnaires (green (catageory 1), yellow (category 2)
and pink (category 3)). The questions involved in the
study are only answered by individuals who were faced with
the green questionnaire. Therefore we keep only these in the
sample.

<<The ethnic minority sample 1,keep.source=FALSE>>=
# table(U$ethnic,exclude=c())
# table(U$s6,exclude=c())
# table(U$a1e,exclude=c())
@
<<The ethnic minority sample 2,>>=
FNSEM <- FNSEM[FNSEM$ethnic!="white" &
     !is.na(FNSEM$ethnic),]
@
<<The ethnic minority sample 3,echo=FALSE>>=
write(nrow(FNSEM),file="crossreference1.tex")
@
<<The ethnic minority sample 4,>>=
FNSEM <- FNSEM[FNSEM$s6!=2,]
FNSEM <- FNSEM[FNSEM$a1e!=3,]
U <- FNSEM <- FNSEM[FNSEM$question==1,]
@
<<The ethnic minority sample 5,keep.source=FALSE,>>=
# table(U$ethnic,exclude=c())
# table(U$s6,exclude=c())
# table(U$a1e,exclude=c())
@

An issue where \citet{Bisin08} is imprecise is whether the
questions they address regard Muslims/non-Muslims or
Muslim/non-Muslim immigrants. Different sample selections
are possible here. The model specifications
in \citet{Bisin08} implies that White Muslims
are excluded and native ethnic minority Muslims are
included. This sample definition does not match
\citet{Bisin08} writing using the terms \emph{Muslim and
non-Muslim immigrants}, as the sample includes natives.

\subsection{Recoding of missing values}

The data set contains several codes for missing values.
These missing values can be of different characters: e.g.,
non--availables, `can't say' or because the
respondent was filtered in a previous filter question. We
employ the strategy to code all these as non--availables in
\proglang{R}; i.e. code \code{NA}. In the data set genuine
non-availables are generally coded as $-1$. We set all $-1$
to \code{NA} in the entire data set:

<<Recoding of missing values>>=
is.na(U) <- U==-1
@

In addition to $-1$ several other codes (\code{na}, $7$,
$8$, $9$, $98$, $99$, $997$ and $999$) are occasionally used
in the data set to indicate various unknown categories. Some
of $-1$ and other unknown categories are non--availables and
have to be deleted in estimations. This is done after we
have coded all our variables in section \ref{sec:subset}. In
questions following a filter question \code{NA} may have to
be set in a category. Some of these other codes used to
denote unknowns are genuine
\code{NA's} and has to be removed. Others will be included
in a category. This is done variable by variable below.

\subsection{Variable definitions}

Using the same variable names as in \citet{Bisin08}, we
define the variables at the precision described by the
authors. We here give our interpretation of the variable
definitions in \citet{Bisin08}.

\subsubsection{Religious affiliation}
Question \code{s6} asks whether the respondent belongs to a
church or has a religion. For those who answer yes,
question \code{s7} asks which that church or religion is.
We define two religious affiliations: \code{muslim},
(which are all who answered category 3 (muslim) on question
\code{s7}) and \code{non-muslim} (all who did not answer
category 3 (muslim) on question \code{s7}). All
non--religious, that is those who answered category 2 on
question \code{s6} are already removed from the sample.
Observations containing \code{na} are recoded to \code{NA}.
\citep[p. 112f]{FNSEM}
<<Religious Affiliation 1,keep.source=FALSE>>=
# table(U$s7,exclude=c())
@
<<Religious Affiliation 2>>=
is.na(U$s7) <- U$s7=="na"
U$Religion  <- ifelse(U$s7=="muslim","muslim","non-muslim")
@
<<Religious Affiliation 3,keep.source=FALSE>>=
# table(U$Religion,exclude=c())
@


\subsubsection{Importance of religion}
Question \code{s9} is about the importance of religion. To
grade the \emph{Importance of Religion} respondents have
to choose between the following categories: 1. not at all
important, 2. not very important, 3. fairly important, or
4. very important. Following standard coding practice of
such questions, $1$ and $2$ should be one  category and 3
and 4 another, but \citet{Bisin08} choose  to put $1$, $2$
and $3$ in the same category.

Among those who have answered the question, very
few have chosen the alternative  $1$ or $2$ in their answer,
implying very skewed distribution. Following
\citet{Bisin08}, code $4$ (`Very important') as answer on
question \code{s9} is coded TRUE; else FALSE. Codes $8$
(`Can't say') and $9$ are coded as \code{NA}. \citep[p.
112f]{FNSEM}

<<Importance of religion 1,keep.source=FALSE>>=
# table(U$s9,exclude=c())
@
<<Importance of religion 2>>=
is.na(U$s9) <-  U$s9==8 | U$s9==9
U$Importance.of.Religion <- U$s9 == 4
@
<<Importance of religion 3,keep.source=FALSE>>=
# table(U$Importance.of.Religion,exclude=c())
@

%</techreport>

%<*present>

\begin{slide}[toc=Two important variables]{Religious
affiliation vs. importance of religion}
\vspace{\stretch{1}}

<<Rowtable function, echo=FALSE>>=
RowTable <- function(x1,x2)  round(cbind(table(x1,x2,exclude=c()),table(x1,exclude=c()))/c(table(x1, exclude=c())),2)
@
\footnotesize{
<<Religion vs Importance of Religion, echo=TRUE>>=
RowTable(U$s7,U$s9)
@
}
\vspace{\stretch{1}}
\end{slide}
%</present>

%<*techreport>
\subsubsection{Attitude towards inter-marriage}

Question \code{s34a} is ``Would you personally mind if a
close relative were to marry a white person?''. It serves
as a filter question to \code{s34b} (``Would you mind very
much or just a little?'') which is asked to those who
answered yes (cod $1$) on \code{s34a}. We code those who
answer yes on both questions (Mind \& Mind very much) as
TRUE. The category FALSE refers then to those who (Do not
mind) or (Mind \& Mind Little). Code $8$ (Can't say) on
\code{s34a} and code $9$  on \code{s34a} and on \code{s34b}
are assigned as \code{NA}. \citep[p. 125]{FNSEM}

<<Attitude towards inter-marriage 1,keep.source=FALSE>>=
# table(U$s34a,exclude=c())
# table(U$s34b,exclude=c())
@
<<Attitude towards inter-marriage 2>>=
is.na(U$s34a) <- U$s34a==8 | U$s34a==9
is.na(U$s34b) <- U$s34b==8
U$Attitude.Towards.Inter.Marriage <-
     U$s34a==1 & U$s34b==1
@
<<Attitude towards inter-marriage 3,keep.source=FALSE>>=
# table(U$Attitude.Towards.Inter.Marriage,exclude=c())
@

%</techreport>

%<*present>

\begin{slide}{Attitude towards inter-marriage}
\vspace{\stretch{1}}
\footnotesize{
What you mind very much if \dots
<<Attitudes, echo=TRUE>>=
RowTable(U$s7,U$Attitude.Towards.Inter.Marriage)
@
}
\vspace{\stretch{1}}
\end{slide}
%</present>

%<*techreport>

\subsubsection{Importance of racial composition in schools}

Two questions are asked: \code{s23} is
\begin{quote}
``If you were choosing a school for an eleven-year old
child of yours, would your choice be influenced by how
many (RESPONDENT'S ETHNIC ORIGIN) children there were in
the school?'' \citep[p. 120]{FNSEM}
\end{quote}
and \code{s24a} asks that if the available school were
similar in other ways would you prefer to send this child
to school with fewer than half of the pupils (code $1$),
about half of the students (code $2$), more than half
(code $3$) were of your own ethnic origin. \code{s23} is
not a filter question. \emph{Importance of Racial
Composition in Schools} is set to TRUE if \code{s24a} is
equal to $3$ and FALSE otherwise. Code $7$ (No preference)
is coded as FALSE. Codes $8$ (Can't say) and $9$  are
assigned as \code{NA}. \citep[p. 120]{FNSEM}

<<Importance of racial composition in schools 1,keep.source=FALSE>>=
# table(U$s24a,exclude=c())
@
<<Importance of racial composition in schools 2>>=
is.na(U$s24a) <- U$s24a==8 | U$s24a==9
U$Importance.of.Racial.Composition.in.Schools <- U$s24a==3
@
<<Importance of racial composition in schools 3,keep.source=FALSE>>=
# table(U$Importance.of.Racial.Composition.in.Schools,
#      exclude=c())
@
%</techreport>

%<*present>

\begin{slide}{Importance of racial composition in schools}
\vspace{\stretch{1}}
\footnotesize{
<<Importance, echo=TRUE>>=
RowTable(U$Religion,
     U$Importance.of.Racial.Composition.in.Schools)
@
}
\vspace{\stretch{1}}
\end{slide}



\begin{slide}[toc=]{}

\end{slide}

\begin{slide}[toc=]{}

\end{slide}

\begin{slide}[toc=]{}

\end{slide}

\begin{slide}[toc=]{}

\end{slide}

\begin{slide}[toc=]{}

\end{slide}

%</present>


%<*techreport>
\subsubsection{Born in the UK}

Defines who is born in the United Kingdom (question
\code{a3}). Category 16 is Northern Ireland, category 17
England and Wales and category 18 Scotland. Code $99$  is
assigned as \code{NA}. \citep[p. 107]{FNSEM}.

<<Born in the UK 1,keep.source=FALSE>>=
#table(U$a3,exclude=c())
@
<<Born in the UK 2>>=
is.na(U$a3)<- U$a3==99
U$Born.in.the.UK <- U$a3==16 | U$a3==17 | U$a3==18
@
<<Born in the UK 3,keep.source=FALSE>>=
#table(U$Born.in.the.UK,exclude=c())
@

\subsubsection{Age at and years since arrival}

This part defines the variables \emph{Age at arrival} and
\emph{Years since arrival} by using information about year
of migration (question \code{a4n}), age (question
\code{a1an}) and the year (variable \code{year} when the
interview is made). The interview is made in \code{93} or
\code{94}). The result is that some individuals get the age
at arrival $-1$, which presumably is due to rounding of
years since arrival and the age.

All born in the UK are coded as $0$ for
\code{Age.at.Arrival} and \code{Years.Since Arrival}. Since
these two variables are related to the age of the
immigrants, one could also add an interaction variable
between \code{Born.in.the.UK} and \code{Age} to account for
effect of age for the natives. The effect of age for natives
is not represented in the model as specified by
\citet{Bisin08}. We experimented with this and results were
basically unchanged. The interaction variable is
insignificant in all specifications. Code $99$  is assigned
as \code{NA} for \code{year}. \citep[pp. 105, 107]{FNSEM}.

<<Age at and years since arrival 1,keep.source=FALSE>>=
# table(U$a4n,exclude=c())
# table(U$a1an,exclude=c())
# table(U$year,exclude=c())
@
<<Age at and years since arrival 2>>=
is.na(U$a4n) <- U$a4n==98 | U$a4n==99
is.na(U$year)<- U$year==99
U$Age <- U$a1an
U$Years.Since.Arrival <- ifelse(U$Born.in.the.UK==TRUE,
     0, U$year-U$a4n)
U$Age.at.Arrival <- U$Age - U$Years.Since.Arrival
U$Age.at.Arrival <- replace(U$Age.at.Arrival,
     U$Born.in.the.UK==TRUE, 0)
@
<<Age at and years since arrival 3,keep.source=FALSE>>=
# table(U$Years.Since.Arrival,exclude=c())
# table(U$Age.at.Arrival,exclude=c())
@

\subsubsection{Female}
Definition of females via question \code{hh2a.s}. Code
\code{na} is coded as \code{NA}. \citep[pp. 318]{FNSEM}.

<<Female 1,keep.source=FALSE>>=
# table(U$hh2a.s,exclude=c())
@
<<Female 2>>=
is.na(U$hh2a.s) <- U$hh2a.s =="na"
U$Female <- U$hh2a=="female"
@
<<Female 3,keep.source=FALSE>>=
# table(U$Female,exclude=c())
@

\subsubsection{Arranged marriage}
\label{sec:Arranged.Marriage}
In question \code{s39} Indian, Pakistani, Bangladeshi and
Chinese respondents who has ever been married were asked a
question about the decision regarding their marriage. The
question ask about the role of the respondent and his or her
parents about choosing the respondent's husband/wife. In
categories $1$ and $2$ of \code{s39} the respondent's parents
made the final decision and these categories define the
dummy where the respondent is or has been living in an
arranged marriage (code $1$; all other are coded $0$).
Notice that singles and Caribbeans have not received this
question. Singles are already removed from the sample (see
section \ref{sec:definingsample}. Caribbeans are coded $0$.
This means that the Caribbeans do not marry according to the
decision of their parents. Category $8$ (Can't say) and
category $9$ are assigned \code{NA}. \citep[pp. 127]{FNSEM}.
<<Arranged marriage 1,keep.source=FALSE>>=
# table(U$s39,exclude=c())
@
<<Arranged marriage 2>>=
is.na(U$s39) <- U$s39==8 | U$s39==9
U$Arranged.Marriage <- ifelse(
     U$s39==3 | U$s39==4 | U$s39==5 |
     U$ethnic=="caribbean", FALSE , TRUE)
@
<<Arranged marriage 3,keep.source=FALSE>>=
# table(U$Arranged.Marriage,exclude=c())
@

\subsubsection{Discrimination}

The \emph{discrimination} variable is based on a series of
questions related to discrimination; \code{v1a}-\code{v1d}
about physical attacks, \code {v9a} about insults,
\code{j55a} and \code{j63a} discrimination at work.
Basically. anyone answering that they have been
discriminated for any of these reasons are coded
$1$; else code $0$.

Questions \code{v1a}-\code{v1d} is a series of filter
questions: Question \code{v1a} asks if the respondent have
been attacked (yes or no), question \code{v1b} how many
attacks the respondent has been enduring and question,
\code{v1c} asks those who have been attacked once if they
believe the attack had to do with reasons to do with race or
colour and \code{v1d} asks the same question and
regards those who have been attacked more than once.
Generally code $8$ and code $9$ are assigned \code{NA},
except for question
\code{v1b} where also code $7$ is assigned \code{NA}.
\citep[pp. 154ff, 163, 195 and 199]{FNSEM}.
<<Discrimination 1>>=
vjlist <- c(paste("v1",letters[1:4],sep=""),
     "v9a","j55a","j63a")
@

<<Discrimination 3>>=
is.na(U[vjlist]) <- U[vjlist]==8 |
     U[vjlist]==9
is.na(U$v1b)  <- U$v1b==7 |  U$v1b==8 |
     U$v1b==9
U$Discrimination   <-
     (U$v1a==1 &  U$v1b==1 &  U$v1c==1) |
     (U$v1a==1 &
     (U$v1b >= 2 & U$v1b <= 6)  &
     U$v1d==1) | U$v9a==1 |  U$j55a==1 |
     U$j63a==1
@
<<Discrimination 4,keep.source=FALSE>>=
# table(U$Discrimination,exclude=c())
@

\subsubsection{Children}

No question about the number of children is asked. Instead
the number of children has to be calculated indirectly via
the number of children not living at home (questions
\code{f16a} and \code{f16b1n}-\code{f16b3n}) and the
relation between the respondent and other persons living in
the household (questions \code{hh2c.b}-\code{hh2.c.m}).

The number of children out of home is calculated in the
following way: If children out of home is TRUE
(\code{f16a=1}), then the  number of children equals the sum
of \code{f16b1n}, \code{f16b2n} and \code{f16b3n} (the
number of children not living at home below 5 years, between
5 and 15 years and above 15 years of age). Else, if there
are no children out of home  (i.e., if \code{f16a=2}), then
the number of children  out of home is set to $0$.
Missing values are coded as below. \citep[p. 57]{FNSEM}.

<<Children 1,keep.source=FALSE>>=
# table(U$f16a,exclude=c())
# table(U$f16b1n,exclude=c())
# table(U$f16b2n,exclude=c())
# table(U$f16b3n,exclude=c())
@
<<Children 2>>=
is.na(U$f16a)    <- U$f16a==8 | U$f16a==9
is.na(U$f16b1n)  <- U$f16b1n==99
is.na(U$f16b2n)  <- U$f16b2n==99
is.na(U$f16b3n)  <- U$f16b3n==98 |
     U$f16b3n==99
U$Child.not.at.Home <- ifelse(U$f16a==1,
     U$f16b1n+U$f16b2n+U$f16b3n,0)
@

Questions \code{hh2c.b}-\code{hh2.c.m} are about the
relationship between the respondent and other individuals in
the household (person b, c, d etc to person m); category $5$
being child of the respondent.  First we check if the
person is a child to the respondent and then all children
are summed over the respondents household adding the
variable measuring number of children not at home.
Generally codes $98$ and  $99$ are assigned \code{NA}.
\citep[p. 318]{FNSEM}.

<<Children 3>>=
hhlist <- c(paste("hh2c.", letters[2:13], sep=""))
@


<<Children 4>>=
is.na(U[hhlist]) <- U[hhlist]==98 | U[hhlist]==99
U$Children <-
     U$Child.not.at.Home + apply(apply(subset(U,
     select=c(hh2c.b:hh2c.m)), 2,
     function(x) x==5),1, function(x) sum(x,
     na.rm=TRUE))
@
<<Children 5,keep.source=FALSE>>=
# table(U$Children,exclude=c())
@
\subsubsection{No British education}
Question \code{q1} asks whether the respondent has any
British education. Code $2$ is no. Code $8$ (Can´t say) is
kept in the alternative category since these individuals
will answer the question \code{q3} about foreign education.
Code $9$ in \code{q1} is assigned \code{NA}. \citep[p.
96]{FNSEM}.
<<No British Education 1,keep.source=FALSE>>=
# table(U$q1,exclude=c())
@
<<No British Education 2>>=
is.na(U$q1) <- U$q1==9
U$No.British.Education <- U$q1==2
@
<<No British Education 3,keep.source=FALSE>>=
# table(U$No.British.Education,exclude=c())
@

\subsubsection{British basic education}

We could not exactly see how this variable was defined in
\citet{Bisin08}. They define the British high education as
A--level and above. One interpretation is then that O-level
are educations included in the basic level. This
interpretation is implemented here. \code{NA} is assigned to
all observations for which the filter question
\emph{No British Education} was \code{NA}.
\citep[pp. 96ff]{FNSEM}.

<<British basic education 1>>=
q2alist <- c(paste("..q2a", c(1:8,12:18), sep=""))
@
<<British basic education 2,keep.source=FALSE>>=
# apply(U[q2alist], 2, function(x) table(x,
#      exclude=c()))
@
<<British basic education 3>>=
U$British.Basic.Education <-
     apply(apply(U[q2alist] ,2,function(x)
     x==1),1, function(x) sum(x, na.rm=TRUE))!=0
U$British.Basic.Education <- ifelse(
     is.na(U$No.British.Education),
     NA,U$British.Basic.Education)
@
<<British basic education 4,keep.source=FALSE>>=
# table(U$British.Basic.Education,exclude=c())
@

\subsubsection{British higher education}
\citet{Bisin08} explicitly defined British higher education
as A-level. Given the definition of British Basic education,
the reference group will include trade apprenticeships as
well as university educations. \code{NA} is assigned to all
observations for which the filter question
\emph{No British Education} was \code{NA}.
 \citep[pp. 96ff]{FNSEM}.

<<British higher education 1,keep.source=FALSE>>=
# table(U$..q2a9,exclude=c())
# table(U$..q2a10,exclude=c())
# table(U$..q2a11,exclude=c())
@
<<British higher education 2>>=
U$British.Higher.Education <- apply(apply(subset(
     U,select=c(..q2a9:..q2a11,..q2a19,..q2a20))
     ,2,function(x) x==1),1,
     function(x) sum(x, na.rm=TRUE))!=0
U$British.Higher.Education <- ifelse(
     is.na(U$No.British.Education),
     NA,U$British.Higher.Education)
@
<<British higher education 3,keep.source=FALSE>>=
# table(U$British.Higher.Education,exclude=c())
@

\subsubsection{Foreign education}

Foreign educations is question \code{q3} asked to all who
answered `no' or `Can't' say' on question \code{q1}.
The answer yes is coded as (code $1$) and no is coded
as(code $2$). Contrary to the above
educational variables \code{NA} is \emph{not} assigned to
all \code{NA} on \emph{No British Education} since some of
them (code $8$) actually was asked the question \code{q3}.
Instead \code{NA} is assigned to all observations for which
the filter question \code{q1} was $1$ or $9$ and to all code
$8$ (Can't say) and code $9$ on question \code{q3}. \citep[p.
99]{FNSEM}.

<<Foreign education 1,keep.source=FALSE>>=
# table(U$q3,exclude=c())
@
<<Foreign education 2>>=
is.na(U$q3) <- U$q3==8 | U$q3==9
U$Foreign.Education <- U$q3==1
@
<<Foreign education 3,keep.source=FALSE>>=
# table(U$Foreign.Education,exclude=c())
@

\subsubsection{Labour market status}

We code the labour market status using \code{j1b} (in paid
work last week or not) and \code{j3occ} (classification of
activity; either last week's activity or potential activity
during the last ten years). The variable \code{j1b} takes
the value $1$ for paid work last week and $2$ otherwise. The
variable \code{j3occ} is coded as follows: \citep[the
former pp. 81f]{FNSEM,FNSEM93b}
\vspace{6pt}
\begin{compactenum}
\item Self-employed (25+ employees)
\item Self-employed (1-24 employees)
\item Self-employed (no employees)
\item Self-employed (employees not known)
\item Manager (establishment of 25+ employees)
\item Manager (establishment of 1-24 employees)
\item Manager (employees not known)
\item Foreman/supervisor
\item Other employee
\item Employee status unknown
\item Not known/not answered
\end{compactenum}
\vspace{6pt}

\paragraph{Employee}
In order to be classified as an employee the individual has
to have answered yes (value $1$) in \code{j1b} and be
classified as employee in \code{j3occ} (value $9$) and
have the value \code{NotAssigned}.\code{NA} is assigned to
categories $10$ and $11$ in \code{j3occ}.

<<Labour market status 1,keep.source=FALSE>>=
# table(U$j1b,exclude=c())
# table(U$j3occ,exclude=c())
@
<<Labour market status 2>>=
is.na(U$j3occ) <- U$j3occ==10 | U$j3occ==11
U$Labour.Market.Status <- ifelse(U$j1b==1 &
     U$j3occ==9 & !is.na(U$j1b==1 &
     U$j3occ==9), "Employee","NotAssigned")
@

\paragraph{Self Employed}

Self--employed are also coded using \code{j1b} and
\code{j3occ};  above. Categories $1-4$ in \code{j3occ} are
defined as self-employed. We also require that \code{j1b} is
equal $1$ (Self--Employed).


<<Labour market status 3>>=
U$Labour.Market.Status <- replace(
     U$Labour.Market.Status,
     (U$j3occ==1 | U$j3occ==2 |
     U$j3occ==3  | U$j3occ==4) &
     U$j1b==1,"SelfEmployed")
@


\paragraph{Manager}
Managers are also coded using \code{j1b} and \code{j3occ};
see above. Categories $5-8$ in \code{j3occ} are defined as
managers (including supervisors).

<<Labour market status 4>>=
U$Labour.Market.Status <-
     replace(U$Labour.Market.Status,
     U$j1b==1 & U$j3occ>4 & U$j3occ<9,
     "Manager")
@

\paragraph{Unemployed}
The question \code{hh5d.s} describes the respondent's labour
market status. Unemployment is defined via this variable.
\code{hh5d.s} is coded in the following way
\citep{FNSEM93b}:
\vspace{6pt}
\begin{compactenum}
\item Full-time education
\item Govt. training programme
\item Full-time paid work
\item Part-time paid work
\item Waiting to take up paid work
\item Registered unemployed
\item Unemployed, not registered
\item Permanently sick or disabled
\item Wholly retired from work
\item Looking after the home
\item Doing something else
\item NA
\end{compactenum}
\vspace{6pt}
We define unemployed as category $6$ and $7$. \citep[p.
324]{Bisin08}. There are few cases where the individual is
classified as Employee according to our definition above and
is reported to be unemployed in \code{hh5d.s}. This can for
example be part-time unemployment. We classify these
individuals as having \code{Unclear} labour market status. These
will be checked later and be removed if they are few in the
final sample.

<<Labour market status 5>>=
U$Labour.Market.Status <-
     replace(U$Labour.Market.Status,
     (U$hh5d.s==6 | U$hh5d.s == 7) &
     U$Labour.Market.Status!="NotAssigned",
     "Unclear")
U$Labour.Market.Status <-
     replace(U$Labour.Market.Status,
     (U$hh5d.s==6 | U$hh5d.s == 7) &
     U$Labour.Market.Status=="NotAssigned",
     "Unemployed")
@

\paragraph{Out of labour force}

The category out of labour force is defined as those having
values (1,2,5,8,9,10,11) in \code{hh5d.s} or value $2$ in
\code{j1b} or \code{j2}.
<<Labour market status 6>>=
U$Labour.Market.Status <-
     replace(U$Labour.Market.Status,
     (U$hh5d.s==1 | U$hh5d.s == 2 |
     U$hh5d.s==5  | U$hh5d.s == 8 |
     U$hh5d.s==9  | U$hh5d.s == 10|
     U$hh5d.s==11 | U$j1b    == 2 |
     U$j2    == 2 ) &
     U$Labour.Market.Status=="NotAssigned",
     "OutOfLabourForce")
@

Remaining observations with the value \code{NotAssigned} in
\code{Labour Market.Status} will be assigned \code{NA}.
At this point there are
\Sexpr{table(U$Labour.Market.Status)[["Unclear"]]}
observations coded as \code{Unclear}. These are now recoded
as \code{NA}.

<<Labour market status 8>>=
is.na(U$Labour.Market.Status) <-
     U$Labour.Market.Status == "NotAssigned"
is.na(U$Labour.Market.Status) <-
     U$Labour.Market.Status == "Unclear"
@

We will create dummy variables using this variable before we
run our models.

<<Labour market status 9,keep.source=FALSE>>=
# table(U$Labour.Market.Status,exclude=c())
@

\subsubsection{No parents}

Variable \emph{No parents} means that the respondent is not
living with his or her parents. This variable is coded with
question \code{f14a} (which takes code $1$ for both alive,
code $2$ father alive, code $3$ for mother alive and $8$ for
both dead) and \code{f14b} (which takes code $2$ if both
living parents do not live with the respondent and code $6$
if the only living parent does not live with the respondent;
else it takes one of the values $1$ or $3-5$).
\emph{No parents} should be coded TRUE if either both
parents are dead or the respondent does not live with any
living parents. However,
since we follow \citet{Bisin08}, we code this variable
including only those who have both their parents dead or
both paprents live away from the respondent. This definition
implies that those who have a parent living away and one
parent dead are assigned the value FALSE. Code $8$ and $9$
are assigned \code{NA} \citep[p. 56]{FNSEM}
<<No parents 1,keep.source=FALSE>>=
# table(U$f14a,exclude=c())
# table(U$f14b,exclude=c())
@
<<No parents 2>>=
is.na(U$f14a) <- U$f14a==9
is.na(U$f14b) <- U$f14b==9
U$No.Parents   <- U$f14a==8 | U$f14b==2
@
<<No parents 3,keep.source=FALSE>>=
# table(U$No.Parents,exclude=c())
@

\subsubsection{Contacts with parents}

The three variables measuring contacts with parents are
defined via three questions asking about the number of
physical contacts (question \code{f15bn}), the number of
contacts via telephone (question \code{f15cn}) and the
number of contacts via letters (question \code{f15dn}) that
the respondent has had with his or her parents during the
last four weeks conditional on not both parents being dead.
All three takes the value of the underlaying variable if at
least one parent is alive and the value $0$ if both parents
are dead. Code $999$ on all three variables and code $997$
on \code{f15bn} are assigned \code{NA}. \citep[p. 56]{FNSEM}

<<Contacts with parents 1,keep.source=FALSE>>=
# table(U$f15bn,exclude=c())
# table(U$f15cn,exclude=c())
# table(U$f15dn,exclude=c())
@
<<Contacts with parents 2>>=
is.na(U$f15bn) <- U$f15bn==997 | U$f15bn==999
is.na(U$f15cn) <- U$f15cn==999
is.na(U$f15dn) <- U$f15dn==999
U$Parents.Physical.Contacts <- ifelse(
     U$f14a!=8,  U$f15bn,0)
U$Parents.Telephone.Calls <- ifelse(
     U$f14a!=8, U$f15cn,0)
U$Parents.Letters <- ifelse(
     U$f14a!=8, U$f15dn,0)
@
<<Contacts with parents 3,keep.source=FALSE>>=
# table(U$Parents.Physical.Contacts,exclude=c())
# table(U$Parents.Telephone.Calls,exclude=c())
# table(U$Parents.Letters,exclude=c())
@


\subsubsection{English language}

There are several language  variables measuring whether the
respondent is speaking English with different individuals;
at home with older, at home with younger, at work and with
friends. We construct theses variables  using question
\code{s12a} which asks whether the respondent regularly
speak to anyone in Britain in any other language than
English and \code{..s12f}, \code{..s12g}, \code{..s12h}  and
\code{..s12i} which asks which language is spoken to the
above mentioned categories of individuals. Each of
\code{..s12f}, \code{..s12g}, \code{..s12h}  and
\code{..s12i} comes in 18 versions (e.g.,
\code{..s12f1},\dots,\code{..s12f18}) where each question
1--15 is coded \code{yes} if the respondent speaks the
language. Question 16 is ``Never speaks to these people/Not
ap[plicable]'', question 17 NA and question 18 `` None of
the above answered positive''. Question 1 is always
regarding English.

The respondent is coded as English speaker if either
\code{s12a} is answered negatively or \code{..s12aX1}, where
\code{X=f,g,h,i}, is answered positively. Only respondents
for which either \code{s12a} is \code{NA} or all of
\code{..s12X1},\dots,\code{..s12X16} are answered negatively
are coded as \code{NA}. Below is the code for \emph{English
Spoken At Home With Older}:
<<English language 1,keep.source=FALSE>>=
# table(U$s12a,exclude=c())
# table(U$..s12f1,exclude=c())
# table(U$..s12g1,exclude=c())
# table(U$..s12h1,exclude=c())
# table(U$..s12i1,exclude=c())
@
<<English language 2>>=
is.na(U$s12a) <- U$s12a== 8 | U$s12a == 9
s12flist <- c(paste("..s12f", 2:16, sep=""))
U$oOLD <- apply(U[s12flist]=="yes", 1, sum)
U$English.Spoken.at.Home.with.Older <-
     ((U$s12a==1 | is.na(U$s12a)) &
     U$..s12f1=="yes") | U$s12a==2
is.na(U$English.Spoken.at.Home.with.Older) <-
     U$oOLD==0 &
     U$English.Spoken.at.Home.with.Older==FALSE
U$English.Spoken.at.Home.with.Older <-
    replace(U$English.Spoken.at.Home.with.Older,
    U$oOLD>0 &
    is.na(U$English.Spoken.at.Home.with.Older),FALSE)
U$DO.NOT.SPEAK.WITH.OLDER <- ifelse(U$..s12f16=="no",0,1)
@

The codes for \emph{English Spoken At Home With Younger},
\emph{English Spoken At Work } and \emph{English Spoken With
Friends} are equivalent. These codes are in
\code{araietal\_source.Rnw} but not shown here. It is also
available in \code{araietal\_source.R}.

<<English language 3,echo=FALSE,keep.source=FALSE>>=
s12glist <- c(paste("..s12g", 2:16, sep=""))
U$oYOUNG <- apply(U[s12glist]=="yes", 1, sum)
U$English.Spoken.at.Home.with.Younger <-
     ((U$s12a==1 | is.na(U$s12a)) &
     U$..s12g1=="yes") | U$s12a==2
is.na(U$English.Spoken.at.Home.with.Younger) <-
     U$oYOUNG==0 &
     U$English.Spoken.at.Home.with.Younger==FALSE
U$English.Spoken.at.Home.with.Younger <-
     replace(U$English.Spoken.at.Home.with.Younger,
     U$oYOUNG>0 &
     is.na(U$English.Spoken.at.Home.with.Younger),
     FALSE)
U$DO.NOT.SPEAK.WITH.YOUNGER <- ifelse(U$..s12g16=="no",0,1)
@

<<English language 4,echo=FALSE,keep.source=FALSE>>=
s12hlist <- c(paste("..s12h", 2:16, sep=""))
U$oWORK <- apply(U[s12hlist]=="yes", 1, sum)
U$English.Spoken.at.Work <-
     ((U$s12a==1 | is.na(U$s12a)) &
     U$..s12h1=="yes") | U$s12a==2
is.na(U$English.Spoken.at.Work) <-
     U$oWORK==0 & U$English.Spoken.at.Work==FALSE
U$English.Spoken.at.Work <-
     replace(U$English.Spoken.at.Work,
     U$oWORK>0 & is.na(U$English.Spoken.at.Work),
     FALSE)
U$DO.NOT.SPEAK.AT.WORK <- ifelse(U$..s12h16=="no",0,1)
@
<<English language 5,echo=FALSE,keep.source=FALSE>>=
s12ilist <- c(paste("..s12i", 2:16, sep=""))
U$oFRIENDS <- apply(U[s12ilist]=="yes",1,sum)
U$English.Spoken.With.Friends <-
     ((U$s12a==1 | is.na(U$s12a)) &
     U$..s12i1=="yes") | U$s12a==2
is.na(U$English.Spoken.With.Friends) <-
     U$oFRIENDS==0 &
     U$English.Spoken.With.Friends==FALSE
U$English.Spoken.With.Friends <-
     replace(U$English.Spoken.With.Friends,
     U$oFRIENDS>0 &
     is.na(U$English.Spoken.With.Friends),  FALSE)
U$DO.NOT.SPEAK.WITH.FRIENDS <- ifelse(U$..s12i16=="no",0,1)
@
<<English language 6,keep.source=FALSE>>=
# table(U$English.Spoken.at.Home.with.Older,
#     exclude=c())
# table(U$English.Spoken.at.Home.with.Younger,
#     exclude=c())
# table(U$English.Spoken.at.Work,exclude=c())
# table(U$English.Spoken.With.Friends,exclude=c())
@



\subsubsection{Household income}

The question \code{hh40} provides information in which
interval the household income of the respondent's household
is. We assign the midpoints in these intervals as the
household income. For the lowest bracket this income is the
midpoint of $[0,77]$. For the highest bracket we assign the
income which is the lowest income in the bracket plus the
income interval down to the midpint of the second highest
bracket; i.e.,
$789+\frac{788-731}2=\Sexpr{round(789+(788-731)/2,1)}$. This
method underestimates the range of the highest bracket but
to a lesser extent than the lower limit $789$.
<<Household income 1,keep.source=FALSE>>=
# table(U$hh40,exclude=c())
@
<<Household income 2>>=
is.na(U$hh40) <- {U$hh40=="refused" |
     U$hh40=="can't say"  | U$hh40=="na"}
U$Household.Income <- c(38.5,96.5,135.0,
     173.5,230.5,260.0,318.0,366.0,414.0,
     471.5,529.5,587.0,649.5,702.0,759.5,
     817.5)[U$hh40]
@
<<Household income 3,keep.source=FALSE>>=
# table(U$Household.Income,exclude=c())
@

\subsubsection{Ward variables}

Ward density of own ethnic group is measured by the variable
\code{wown} in the original data. This variable is coded
$1-7$ depending on the density of the respondent's own
ethnic group is in the ward of the respondent
\citep{FNSEM93b}:
\vspace{6pt}
\begin{compactenum}
\item Up to 1.99\%
\item 2-4.99\%
\item 5-9.99\%
\item 10-14.99\%
\item 15-24.99\%
\item 25-32.99\%
\item 33\% or more
\end{compactenum}
\vspace{6pt}
We recode the variable to take the midpoints of the
density intervals in the same fashion as the household
income variable was recoded. This means that if the
respondent is the lowest interval the density is set to be
$1$ etc. In the highest interval we set the density to be
the lowest density in the interval plus the density distance
down to the midpoint of the second highest density interval;
i.e., $33+\frac{33-25}2=\Sexpr{round(33+(33-25)/2,1)}$.

The variable \code{wunemp} is coded $1-6$ depending on
unemployment rate in the ward of the respondent
\citep{FNSEM93b}:
\vspace{6pt}
\begin{compactenum}
\item Up to 1.99\%
\item 2-4.99\%
\item 5-9.99\%
\item 10-14.99\%
\item 15-20\%
\item 20\% or more
\end{compactenum}
\vspace{6pt}
We recode this variable to instead take the midpoints of the
intervals as we did for the household income.
This means that if the respondent is the lowest interval the
the rate is set to be $1$ etc. In the highest interval we
set the rate to be the lowest rate in the interval plus
the distance down to the midpoint of the second
highest rate interval; i.e.,
$20+\frac{20-15}2=\Sexpr{round(20+(20-15)/2,1)}$.

<<Ward variables 1,keep.source=FALSE>>=
# table(U$wown,exclude=c())
@
<<Ward variables 2>>=
U$Ward.Density.Own.Ethnicity <- (c(1.0,3.5,
     7.5,12.5,20.0,29.0,37.0)[U$wown])/100
U$Ward.Unemployment.Rate <- c(1.0,3.5,7.5,12.5,
     17.5,22.5)[U$wunemp]
@
<<Ward variables 3,keep.source=FALSE>>=
# table(U$Ward.Density.Own.Ethnicity,exclude=c())
# table(U$Ward.Unemployment.Rate,exclude=c())
@

\subsection{Discrimination own ethnicity}

Finally we define a variable describing the discrimination
against the own ethnic group. It is defined as the average
of the variable \emph{Discrimination} over ethnic groups
after the removal of non--availables.

<<Another variable definition>>=
GroupDiscrimination <-
     tapply(U$Discrimination,U$ethnic,
              function(x) mean(x, na.rm=TRUE))
U$Discrimination.Own.Ethnicity <-
     GroupDiscrimination[U$ethnic]
@

\subsection{Defining the subset}
\label{sec:subset}

We define dummy variables for labour market status to have
the same variable labels as in \citet{Bisin08}.

<<Defining dummies for Labour Market Status>>=
U$Employee <- as.numeric(U$Labour.Market.Status=="Employee")
U$Manager <- as.numeric(U$Labour.Market.Status=="Manager")
U$Self.Employed <-
      as.numeric(U$Labour.Market.Status=="SelfEmployed")
U$OUT.OF.LABOUR.FORCE <-
      as.numeric(U$Labour.Market.Status=="OutOfLabourForce")
U$Unemployed <-
      as.numeric(U$Labour.Market.Status=="Unemployed")
@

We save a data set keeping all observations containing
non--availables:

<<Defining the subset 1>>=
U.Original <- U
@
We then choose the variables to keep in \code{U}.
This code is in \code{araietal\_source.Rnw} but not shown
here. It is also available in \code{araietal\_source.R}.


<<Defining the subset 2,keep.source=FALSE,echo=FALSE>>=
U <- subset(U, select=c(
    Religion,
    s7,
    ethnic,
    Importance.of.Religion,
    Attitude.Towards.Inter.Marriage,
    Importance.of.Racial.Composition.in.Schools,
    Age.at.Arrival,
    Age,
    Female,
    Born.in.the.UK,
    Arranged.Marriage,
    Discrimination,
    Children,
    Years.Since.Arrival,
    No.British.Education,
    British.Basic.Education,
    British.Higher.Education,
    Foreign.Education,
    Employee,
    Manager,
    Self.Employed,
    OUT.OF.LABOUR.FORCE,
    Unemployed,
    No.Parents,
    Parents.Physical.Contacts,
    Parents.Telephone.Calls,
    Parents.Letters,
    English.Spoken.at.Home.with.Older,
    DO.NOT.SPEAK.WITH.OLDER,
    English.Spoken.at.Home.with.Younger,
    DO.NOT.SPEAK.WITH.YOUNGER,
    English.Spoken.at.Work,
    DO.NOT.SPEAK.AT.WORK,
    English.Spoken.With.Friends,
    DO.NOT.SPEAK.WITH.FRIENDS,
    Household.Income,
    Ward.Density.Own.Ethnicity,
    Ward.Unemployment.Rate,
    Discrimination.Own.Ethnicity,
    area,
    weightis))
@

Finally we remove all observations containing
non--availables from \code{U}:
<<Defining the subset 3>>=
U <- na.omit(U)
@

\subsection{Sample statistics}
%</techreport>
%<*paper>
In our definition of the relevant (ethnic minority) sample
we have excluded (i) the UK majority population (defined as
Whites  in the data set), (ii) all who do not have a
religion or do not belong to a church since they cannot be
classified in a religious group, (iii) all singles since
only people who are married or have been married answer the
question about who made the final decision of their marriage
and (iv) those who answered questionnaire ``yellow'' and
``pink'' since they do not answer relevant questions
involved in the study.
%</paper>
%<*paper|techreport>
Table \ref{ta:diffsample}   compares  the number of
observations in this sample \emph{before}
%</paper|techreport>
%<*techreport>
(see section \ref{sec:subset}; i.e. data frame
\code{U.Original})
%</techreport>
%<*paper|techreport>
and \emph{after}
%</paper|techreport>
%<*techreport>
(see section \ref{sec:subset}; i.e. data frame \code{U})
%</techreport>
%<*paper|techreport>
non--availables are removed with the numbers of observations
reported by the \citet{Bisin08} study. The number of
observations for various groups in the non--Muslim category
are not reported in the \citet{Bisin08} paper. These numbers
are therefore missing in the table. The category definitions
are from the original dataset and involves no recoding on
our part.

After removing observations with missing values on all
variables of interest (``After'' in Table
\ref{ta:diffsample}), we are left with
\Sexpr{nrow(U[U$Religion=="muslim",])} Muslims and
\Sexpr{nrow(U[U$Religion=="non-muslim",])} non-Muslims. The
sample selections induced by the choice of variables and the
missing values in these variables lead to a loss of
$\input{crossreference2.tex}$ percent of the
relevant sample of the original
data.\footnote{The variables written in capital letters
are created to ensure well--defined reference categories.
They are included in our regressions, but we cannot say
whether they are included in the regressions of
\citet{Bisin08}.}


The sample means reported in \citet{Bisin08} seem to be
unweighted. Since data instructions says that the
data should always be weighted,  tables \ref{tab:descriptive}
and \ref{tab:descriptive2} report weighted and unweighted
sample means before and after removal of non-availables and
the \citet{Bisin08} data.\footnote{See \citet{FNSEM93b}.}
Comparing means, the \citet{Bisin08} data seem to be
different from the original sample. The variables
\emph{Attitude Towards Inter Marriage} and
\emph{Importance of Racial Composition in Schools}
in \citet{Bisin08} data deviate largely from corresponding
averages in the original data. The deviation is extreme in
case of \emph{Importance of Racial Composition in Schools}.
 The original sample has a mean for this variable that is
\Sexpr{100*round(mean(U.Original$Importance.of.Racial.Composition.in.Schools[U.Original$Religion=="non-muslim"],na.rm=TRUE),2)}
percent for non-Muslims (compare with 33 percent in
\citet{Bisin08}) and
\Sexpr{100*round(mean(U.Original$Importance.of.Racial.Composition.in.Schools[U.Original$Religion=="muslim"],na.rm=TRUE),2)}
percent for Muslims (compare with 65 percent in
\citet{Bisin08}). Due to this extremely skewed distribution.
it is hardly meaningfull to run a regression on this
variable,

Notice that also the distribution of the variable
\emph{Importance of Religion} would be extremely skewed
using standard coding of this type of variables. Such a
coding would imply that religion is important when the
respondent answer ``Very Important'' and ``Fairly
Important'', to the question ``How important is religion to
the way you live your life?''.


The sample means in our data after removing  accumulated
missing values due to all variables in the estimations
deviate marginally in general from the original data.
The similarities here are partly due to the fact that the
statistics are based exactly on the same variable definition
in our implementation. In some respects, the deviations are
larger. For further comparisons we refer to Tables
\ref{tab:descriptive} and \ref{tab:descriptive2}.

Due to the fact that the large majority of observations
from the original data are lost, the remaining sample is
likely to be contaminated with sample selection bias.
To compare the characteristics of the remaining sample
with the original sample says something about systematic
attrition with respect to observables. The sample selection
bias with respect to unobservables cannot, however,
be resolved.

\section{Regression Results}
\label{Regression}


%</paper|techreport>
<<Defining models for regressions, echo=FALSE>>=
# Defining the regression model for
# Importance of Religion
# DO.NOT.SPEAK.WITH.YOUNGER, removed,
# no observations for Muslims.
# only two observations for non-Muslims
# table(U$DO.NOT.SPEAK.WITH.YOUNGER, U$Religion)

model1 <- {formula(Importance.of.Religion ~
    Age.at.Arrival + Female + Born.in.the.UK +
    Arranged.Marriage + Discrimination +
    Children + Years.Since.Arrival +
    No.British.Education + British.Basic.Education +
    British.Higher.Education + Foreign.Education +
    Employee+ Manager + Self.Employed + Unemployed+
    No.Parents + Parents.Physical.Contacts +
    Parents.Telephone.Calls + Parents.Letters +
    English.Spoken.at.Home.with.Older +
    English.Spoken.at.Home.with.Younger +
    English.Spoken.at.Work +
    English.Spoken.With.Friends +
    Household.Income +
    Discrimination.Own.Ethnicity +
    Ward.Density.Own.Ethnicity +
    Ward.Unemployment.Rate + area +
    factor(DO.NOT.SPEAK.WITH.OLDER)+
    factor(DO.NOT.SPEAK.AT.WORK) +
    factor(DO.NOT.SPEAK.WITH.FRIENDS))}


# Defining the model using the same regressors as
# in Model 1 and Attitude.Towards.Inter.Marriage
# as dependent variable
model2 <- {update(model1,
     Attitude.Towards.Inter.Marriage ~ .)}

# Defining the model using the same regressors as
# in Model 1 and Importance.of.Racial.Composition.in.Schools
# as dependent variable
model3 <- {update(model2,
     Importance.of.Racial.Composition.in.Schools ~ .)}
@

<<Regressions, echo=FALSE>>=
library(sandwich,warn.conflicts=FALSE)
library(lmtest,warn.conflicts=FALSE)
Umuslim  <- U[U$Religion=="muslim",]
UNmuslim <- U[U$Religion!="muslim",]

ESTIMATE <- function(x,df)
       {fm <- lm(x, weights=weightis, data=df)
        rbind(coeftest(fm,
               vcov = vcovHC(fm, type ="HC1")),
                       summary(fm)$adj.r.squared)}

result1.muslim <- ESTIMATE(model1, Umuslim)
result2.muslim <- ESTIMATE(model2, Umuslim)
result3.muslim <- ESTIMATE(model3, Umuslim)


result1.non.muslim <- ESTIMATE(model1, UNmuslim)
result2.non.muslim <- ESTIMATE(model2, UNmuslim)
result3.non.muslim <- ESTIMATE(model3, UNmuslim)

@

%<*paper|techreport>
We use linear probability models
(LPM).\footnote{\citet{Bisin08} use probit estimations. Our
attempts to use probit run into convergence problems. The
convergence problems are severe for the model using
\emph{Importance of Racial Composition in Schools} as
dependent variable. Hence, our choice of LPM.

Another issue is that \citet{Bisin08} should have included
dummy variables indicating religious affiliation:
Christians, Sikhs and others in the non-Muslim category to
check similarities and differences among non-Muslims as
well. In this respect we follow their model specification.

Moreover, \citet{Bisin08} should have adjusted for within
ward correlations. This might matter for their standard
errors, which might be underestimated. In our case, with
almost no significant results, this would not matter much.
The variable is not available in the data set and we did not
make much effort to obtain it.
%</paper|techreport>



%<*techreport>
All estimated models include 7 UK-region dummies, and
the variables \code{DO.NOT.} \code{SPEAK.WITH.OLDER},
\code{DO.NOT.SPEAK.AT.WORK}, and
\code{DO.NOT.SPEAK.WITH.FRIENDS}.  It turned out that the
variable \code{DO.NOT.SPEAK.WITH.YOUNGER} is TRUE for few
observations and cannot be included in the model.
%</techreport>
%<*paper|techreport>
} Our results are
presented in Tables \ref{ta:regressions} and
\ref{ta:regressions2}. \citet{Bisin08} write that:
\begin{enumerate}
\item ``Muslims integrate less and more slowly than
non-Muslims.'' (abstract, p. 445) and \item  ``\dots there
is no evidence that segregated neighbourhoods breed intense
religious and cultural identities. On the contrary, \dots
intense identities in our data are more prominent
in relatively mixed neighbourhoods.''
(p. 446)
\end{enumerate}

The first claim is based on their reported results
concerning the variable \emph{Years Since Arrival}.  In
this way \citet{Bisin08} compare cohorts of Muslims and non
Muslims and attempt to say something about the evolution of
values over time. They do not follow individuals over time
but nonetheless call these cohort differences ``Integration
over time''.  They report negative coefficients for
\emph{Years Since Arrival}, but the estimates are smaller
in absolute value for Muslims than for non-Muslims. In our
case, the coefficients for \emph{Years Since Arrival}
reported  in Tables \ref{ta:regressions} and
\ref{ta:regressions2} are insignificant in all cases except
in the regression for \emph{Importance of Religion} for
Muslims, where it is negative. This is opposite to what
\citet{Bisin08} claim.

The second claim is based on their reported results
concerning the variable \emph{Ward Density Own Ethnicity}.
\citet{Bisin08} report negative and significant estimates
for \emph{Ward Density Own Ethnicity} in all six
specifications. Their negative coefficient for this variable
would imply that ethnic minorities put more weight on
religion, mind more about inter-ethnic marriage and have
stronger taste for ethnically profiled schools, as we move
from neighbourhoods (Wards) with high density of their own
ethnicity to neighbourhoods where people from their own
ethnicity are scarce. This is not at all what we find in our
replication.

In our estimations, the estimated coefficients for this
variable are all positive but far from significant. The
P-values are
\Sexpr{round(result1.muslim["Ward.Density.Own.Ethnicity",4],2)},
\Sexpr{round(result2.muslim["Ward.Density.Own.Ethnicity",4],2)}
 and
\Sexpr{round(result3.muslim["Ward.Density.Own.Ethnicity",4],2)}
for Muslims and
\Sexpr{round(result1.non.muslim["Ward.Density.Own.Ethnicity",4],2)},
\Sexpr{round(result2.non.muslim["Ward.Density.Own.Ethnicity",4],2)}
and
\Sexpr{round(result3.non.muslim["Ward.Density.Own.Ethnicity",4],2)}
for non-Muslims contradicting the \citet{Bisin08} results.

Inspecting the results presented in Tables
\ref{ta:regressions} and \ref{ta:regressions2}, there are
many similarities and few differences in the estimated
coefficients for Muslims and non-Muslims. Our results are
generally very different from results reported by
\citet{Bisin08}. We are, however, doubtful whether it is
possible to draw any reliable inference from these results
due to great loss of observations and possible sample
selection bias, together with the problem of endogeneity
(also mentioned by \citet{Bisin08}).



\section{Concluding remarks}
\label{Discussion}

The \citet{Bisin08} paper rests on fragile grounds. Our
examination of the data using their variable definitions and
the same set-up indicates that their claims about
differences between Muslims and non-Muslims, and their
conclusion that strong Religious/Ethnic identities are found
in mixed neighbourhoods does not hold. There is no
systematic relation between ethnic minorities' views on
religion, inter-ethnic marriage or ethnic profile of schools
and the density of their own ethnic minority in their
neighbourhood. However, we hesitate to draw inference from
these results since the great loss of observations
($\input{crossreference2.tex}$ percent) implies that the
remaining sample is most likely not representative.



%</paper|techreport>

%<*present>
\begin{slide}{Conclusions}
\vspace{\stretch{1}}
\begin{itemize}
\item The \citet{Bisin08} paper rests on fragile grounds.
\item Their claims about differences between Muslims and
non-Muslims, and their conclusion that strong
Religious/Ethnic identities are found in mixed
neighbourhoods does not hold.
\item There is no
systematic relation between ethnic minorities' views on
religion, inter-ethnic marriage or ethnic profile of schools
and the density of their own ethnic minority in their
neighbourhood.
\item However, we hesitate to draw inference from
these results since the great loss of observations
 ($\input{crossreference2.tex}$ percent) implies that the
remaining sample is most likely not representative.
\end{itemize}
\vspace{\stretch{1}}
\end{slide}

%</present>



%<*techreport>
\section{Production notes}
\label{sec:prodnotes}

To facilitate reproducibility and save others timely
interpretations of what is done in this paper, we attempt
to follow Literate Statistical Programming procedures. For
documenting our results we have used \pkg{Sweave} by
\citet{Leisch02} in combination with the \LaTeX\ family of
programs using the packages \code{inputenc}, \code{fontenc},
\code{natbib}, \code{Sweave}, \code{fancyvrb}, \code{color},
\code{url}, \code{hyperref} and \code{multirow}.

All code (\proglang{R} code, \LaTeX\ code and Bib\TeX\ data
base code) used to do the econometric estimations, to
produce this technical documentation including all tables
and to produce the companion paper \citet{Araietal08a} is
contained in the file \code{araietal\_source.Rnw}.

The estimations and the documentation can be reproduced in
by following the instructions at the top of the file
\url{http://people.su.se/~lundh/fragile_grounds/araietal\_source.Rnw}

Our results were obtained on a
\Sexpr{sessionInfo()$R.version$platform} platform
using
\proglang \Sexpr{sessionInfo()$R.version$version.string}
\citep{Rcore} with packages
\pkg{\Sexpr{sessionInfo()$otherPkgs$lmtest$Package}}
     \Sexpr{sessionInfo()$otherPkgs$lmtest$Version}
    (\Sexpr{sessionInfo()$otherPkgs$lmtest$Date}),
\pkg{\Sexpr{sessionInfo()$otherPkgs$sandwich$Package}}
     \Sexpr{sessionInfo()$otherPkgs$sandwich$Version}
    (\Sexpr{sessionInfo()$otherPkgs$sandwich$Date}),
\pkg{\Sexpr{sessionInfo()$otherPkgs$zoo$Package}}
     \Sexpr{sessionInfo()$otherPkgs$zoo$Version}
    (\Sexpr{sessionInfo()$otherPkgs$zoo$Date}),
\pkg{\Sexpr{sessionInfo()$otherPkgs$foreign$Package}}
     \Sexpr{sessionInfo()$otherPkgs$foreign$Version}
    (\Sexpr{sessionInfo()$otherPkgs$foreign$Date}) and
\pkg{\Sexpr{sessionInfo()$otherPkgs$xtable$Package}}
     \Sexpr{sessionInfo()$otherPkgs$xtable$Version}
    (\Sexpr{sessionInfo()$otherPkgs$xtable$Date}).


%</techreport>

%<*paper|techreport>


\bibliography{araietal}

\bibliographystyle{plainnat}

\newpage
\section*{Appendix: Tables}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Comment: We use a combination of xtable() and print()
%% functionality in order to have R produce the desired
%% tables directly i LaTeX-code.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
<<Loading xtable,echo=FALSE>>=
library(xtable)
@

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% Code for Table 1: Religious affiliation ...
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
<<Table 1,results=tex,keep.source=FALSE,echo=FALSE>>=
# Defining variables to use in the table
OnonMuslim        <- table(FNSEM$s7!="muslim")[2]
names(OnonMuslim) <- "nonMuslim"
SnonMuslim        <- table(U$Religion!="muslim")[2]
names(SnonMuslim) <- "nonMuslim"
# Defining the table header
Header <- paste("Religious affiliation",
"&\\multicolumn{1}{c}{(1)}",
     "&\\multicolumn{1}{c}{(2)}",
     "&\\multicolumn{1}{c}{(3)}",
     "&\\multicolumn{1}{c}{(4)}",
     "&\\multicolumn{1}{c}{(5)}",
     "&\\multicolumn{1}{c}{(6)}",
     "\\\\\\hline",
     "&\\multicolumn{2}{c}{Before}",
     "&\\multicolumn{2}{c}{After}",
     "&\\multicolumn{2}{c}{Bisin et al.}",
     "\\\\",
     "&\\multicolumn{2}{c}{$n=",length(U.Original$s7),"$}",
     "&\\multicolumn{2}{c}{$n=",length(U$s7),"$}",
     "&\\multicolumn{2}{c}{$n=5963$}",
     "\\\\\\cline{2-7}",
     "&\\multicolumn{1}{r}{\\#}&\\multicolumn{1}{r}{\\%}",
     "&\\multicolumn{1}{r}{\\#}&\\multicolumn{1}{r}{\\%}",
     "&\\multicolumn{1}{r}{\\#}&\\multicolumn{1}{r}{\\%}\\\\",
     sep="")
# Defining table footer
Footer <- paste("\\hline All non-Muslims",
      "&",OnonMuslim,"&",
      round(100*OnonMuslim/table(FNSEM$s7=="na")[[1]],2),
      "&",SnonMuslim,"&",
      round(100*SnonMuslim/length(U$Religion),2),
      "&3594&",round(100*3594/(3594+2369),2),
      "\\\\\\hline",
      "\\multicolumn{7}{l}{\\multirow{2}{10.5cm}",
      "{\\footnotesize NOTE: The row names shows exactly ",
      "how the original data is coded, so that e.g., ",
      "`NA's' are true missing values whereas `na' is ",
      "coded as religious affiliation `na'. On the last ",
      "line non--Muslims are calculated excluding na and ",
      "NA.}}\\\\",
       sep="")
addtorow <- list()
addtorow$pos <- list()
addtorow$pos[[1]] <- 0
addtorow$pos[[2]] <- 13
addtorow$command <- c(Header,Footer)
# Defining the data to use in the table
DiffSample <- cbind(
     summary(FNSEM$s7),
     100*summary(U.Original$s7)/length(U.Original$s7),
     c(summary(U$s7),
     length(U$s7)-table(is.na(U$s7))[[1]]),
     c(100*summary(U$s7)/length(U$s7),
     100*(length(U$s7)-table(is.na(U$s7))[[1]])),
     matrix(list(c(),c(),round(2369,0),c(),c(),c(),
     c(),c(),c(),c(),c(),c(),c())),
     matrix(list(c(),c(),39.73,c(),c(),c(),c(),c(),
     c(),c(),c(),c(),c()))
     )
# Taking dimnames from original data
dimnames(DiffSample)[[1]] <- names(summary(U.Original$s7))
# Generate the table
TabDiffSample <- print(xtable(DiffSample,align=c("lrrrrrr"),
     caption=c("Religious affiliation (absolute (\\#) and
     relative (\\%) numbers), before (columns 1 and 2) and
     after (columns 3 and 4) removal of \\code{NA} compared
     with \\citet{Bisin08} (columns 5 and 6)."),
     label="ta:diffsample",
     digits=c(0,0,2,0,2,0,2)),
     table.placement = "h",
     caption.placement="top",
     include.colnames=FALSE,
     add.to.row=addtorow,
     hline.after=c(-1,0,11))
@

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% Code for Table 2-3: Weighted and weighted means ...
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

<<Table 2-3,results=tex,keep.source=FALSE,echo=FALSE>>=
# Generate sample with all NA removed
SampleU <- subset(U,select=c(
     Importance.of.Religion,
     Attitude.Towards.Inter.Marriage,
     Importance.of.Racial.Composition.in.Schools,
     Age.at.Arrival,
     Age,
     Female,
     Born.in.the.UK,
     Arranged.Marriage,
     Discrimination,
     Children,
     Years.Since.Arrival,
     No.British.Education,
     British.Basic.Education,
     British.Higher.Education,
     Foreign.Education,
     Employee,
     Manager,
     Self.Employed,
     OUT.OF.LABOUR.FORCE,
     Unemployed,
     No.Parents,
     Parents.Physical.Contacts,
     Parents.Telephone.Calls,
     Parents.Letters,
     English.Spoken.at.Home.with.Older,
     DO.NOT.SPEAK.WITH.OLDER,
     English.Spoken.at.Home.with.Younger,
     DO.NOT.SPEAK.WITH.YOUNGER,
     English.Spoken.at.Work,
     DO.NOT.SPEAK.AT.WORK,
     English.Spoken.With.Friends,
     DO.NOT.SPEAK.WITH.FRIENDS,
     Household.Income,
     Ward.Density.Own.Ethnicity,
     Ward.Unemployment.Rate,
     Discrimination.Own.Ethnicity
))
# Generate sample with no NA removed
SampleU.Original <- subset(U.Original,select=c(
     Importance.of.Religion,
     Attitude.Towards.Inter.Marriage,
     Importance.of.Racial.Composition.in.Schools,
     Age.at.Arrival,
     Age,
     Female,
     Born.in.the.UK,
     Arranged.Marriage,
     Discrimination,
     Children,
     Years.Since.Arrival,
     No.British.Education,
     British.Basic.Education,
     British.Higher.Education,
     Foreign.Education,
     Employee,
     Manager,
     Self.Employed,
     OUT.OF.LABOUR.FORCE,
     Unemployed,
     No.Parents,
     Parents.Physical.Contacts,
     Parents.Telephone.Calls,
     Parents.Letters,
     English.Spoken.at.Home.with.Older,
     DO.NOT.SPEAK.WITH.OLDER,
     English.Spoken.at.Home.with.Younger,
     DO.NOT.SPEAK.WITH.YOUNGER,
     English.Spoken.at.Work,
     DO.NOT.SPEAK.AT.WORK,
     English.Spoken.With.Friends,
     DO.NOT.SPEAK.WITH.FRIENDS,
     Household.Income,
     Ward.Density.Own.Ethnicity,
     Ward.Unemployment.Rate,
     Discrimination.Own.Ethnicity
))

# Set Age.at.Arrival and Years.Since.Arrival
# for those Born.in.the.UK to NA. This is only for the
# sample statistics.


is.na(SampleU.Original$Age.at.Arrival)      <-
                       SampleU.Original$Born.in.the.UK==1
is.na(SampleU.Original$Years.Since.Arrival) <-
                       SampleU.Original$Born.in.the.UK==1


# A function to compute weighted mean
# for a variable in a data set (x), with
# weights (w) and column (n).
WMEAN <- function (x,W,n) {
     weighted.mean(x[!is.na(x[,n]),n], W[!is.na(x[, n])])}


# Weights to use in computing weighted means for Muslims
Muslim.weight <-
     U$weightis[U$Religion=="muslim"]
Orig.Muslim.weight <-
     U.Original$weightis[U.Original$Religion=="muslim"]
Non.Muslim.weight <-
     U$weightis[U$Religion=="non-muslim"]
Orig.Non.Muslim.weight <-
     U.Original$weightis[U.Original$Religion=="non-muslim"]


# Create a matrix to gather means
# Columns 1:5 for Muslims:
# 1:2 weighted means, 3:4 unweighted, 5 Bisin et al.
# Columns 6:10 for non-Muslims:
# 6:7 weighted means, 8:9 unweighted, 10 Bisin et al.

Means.Table <- matrix(,ncol(SampleU),10)
rownames(Means.Table) <- colnames(SampleU)

# Compute weighted means in columns 1:36
for (i in 1:ncol(SampleU.Original)) {Means.Table[i,1]<-
     WMEAN(SampleU.Original[U.Original$Religion=="muslim",],
                                      Orig.Muslim.weight, i)}

# Compute weighted means in columns 1:36
for (i in 1:ncol(SampleU)) {Means.Table[i,2]<-
     WMEAN(SampleU[U$Religion=="muslim",] , Muslim.weight, i)}



# Unweighted Means for Muslims
Means.Table[,3] <-
     mean(SampleU.Original[U.Original$Religion=="muslim",],
     na.rm=TRUE)
Means.Table[,4] <-
     mean(SampleU[U$Religion=="muslim",], na.rm=TRUE)


# Bisin et al. Muslims
Means.Table[,5]  <-
c(0.79,0.70,0.65,39.18, NA, 0.47,0.21,0.22,0.17,2.17,
     26.43,0.81,0.06,0.08,0.25,0.38,0.02,0.09,NA,0.19,
     0.34,3.05,3.38,0.67,0.03,NA,0.20,NA,0.19,NA,0.22,
     NA,200.74,0.15,16.57,0.21)


# Compute weighted means SampleU.Original
for (i in 1:ncol(SampleU.Original)) {Means.Table[i,6]<-
     WMEAN(SampleU.Original[U.Original$Religion=="non-muslim",],
     Orig.Non.Muslim.weight, i)}

# Compute weighted means SampleU
for (i in 1:ncol(SampleU)) {Means.Table[i,7]<-
     WMEAN(SampleU[U$Religion=="non-muslim",],
     Non.Muslim.weight, i)}


# Unweighted Means for non-Muslims
Means.Table[,8] <-
     mean(SampleU.Original[U.Original$Religion=="non-muslim",],
         na.rm=TRUE)
Means.Table[,9] <-
     mean(SampleU[U$Religion=="non-muslim",],na.rm=TRUE)



# Bisin et al. non-Muslims
Means.Table[,10]  <-
c(0.42,0.37,0.33,42.57,NA,0.48,0.28,0.12,0.19,1.68,
     27.08,0.52,0.13,0.16,0.29,0.59,0.04,0.14,NA,0.08,
     0.32,3.87,4.74,0.37,0.08,NA,0.25,NA,0.27,NA,0.27,
     NA,330.26,0.11,12.60,0.18)



# Replace "." with " " in variable names to appear in the table
rownames(Means.Table) <- gsub("\\." , " ", rownames(Means.Table))



# Creating the table i LaTeX code
print(xtable(Means.Table[1:17,],align=c("lrrrrrrrrrr"),
     caption="Weighted and Unweighted Means for Muslims and
     non--Muslims before and after removal of \\code{NA}
     compared with \\citet{Bisin08}.",
     label="tab:descriptive"),
     table.placement = "p",
     floating.environment = "sidewaystable",
     caption.placement="top",
     add.to.row=list(pos=list(0),
     command=paste(
     "&\\multicolumn{5}{c}{Muslim}",
     "&\\multicolumn{5}{c}{Non-Muslim}\\\\\\hline",
     "&\\multicolumn{2}{c}{W e i g h t e d}",
     "&\\multicolumn{3}{c}{U n w e i g h t e d}",
     "&\\multicolumn{2}{c}{W e i g h t e d}",
     "&\\multicolumn{3}{c}{U n w e i g h t e d}\\\\",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Bisin}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Bisin}\\\\",
     "&&&&&et al.&&&&&et al.\\\\\\hline",
     sep="")),
     include.colnames=FALSE,
     hline.after=c(-1,0,NULL))
# Table 3: Continuation of Table 2
print(xtable(Means.Table[18:nrow(Means.Table),],
     align=c("lrrrrrrrrrr"),
     caption="Table 2 continued. Weighted and Unweighted
     Means for Muslims and non--Muslims before and after
     removal of \\code{NA} compared with \\citet{Bisin08}.",
     label="tab:descriptive2"),
     table.placement = "p",
     floating.environment = "sidewaystable",
     caption.placement="top",
     add.to.row=list(pos=list(0),
     command=paste(
     "&\\multicolumn{5}{c}{Muslim}",
     "&\\multicolumn{5}{c}{Non-Muslim}\\\\\\hline",
     "&\\multicolumn{2}{c}{W e i g h t e d}",
     "&\\multicolumn{3}{c}{U n w e i g h t e d}",
     "&\\multicolumn{2}{c}{W e i g h t e d}",
     "&\\multicolumn{3}{c}{U n w e i g h t e d}\\\\",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Bisin}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Before}",
     "&\\multicolumn{1}{c}{After}",
     "&\\multicolumn{1}{c}{Bisin}\\\\",
     "&&&&&et al.&&&&&et al.\\\\\\hline",
     sep="")),
     include.colnames=FALSE,
     hline.after=c(-1,0,NULL))
@
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% Code for Table 4-5: Regression results ...
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

<<Table 4 prepare, echo=FALSE>>=
MakeTable <- function(x, n=2) {
    # take coef() objects and makes table
    Stars <- ifelse(x[,4]<0.05, paste("*"), paste(""))
    names(Stars) <- NULL
    Estimates  <- format(round(x[,1],n))
    Std.errors <- format(round(x[,2],n))
    x <- cbind(Estimates,Std.errors, Stars)
    Table <- matrix(,2*nrow(x),1)
    Table[seq(1,2*nrow(x)-1,2),1] <-
    paste(x[,1],x[,3], sep="")
    Table[seq(2,2*nrow(x),2  ),1] <-
    paste("(",x[,2],")",sep="")
    Xname  <- unlist(strsplit(unlist(strsplit(rownames(x),
    "TRUE")), "yes") )
    Xname  <- gsub("\\.", " ", Xname)
    Xnames <- matrix(,2*nrow(x),1)
    Xnames[seq(1,2*nrow(x)-1,2),1] <- Xname
    Xnames[seq(2,2*nrow(x)  ,2),1] <- ""
    strsplit(Xnames, "TRUE")
    strsplit(Xnames, "yes")
    rownames(Table) <- Xnames
    Table
    }


Table.Reg <-
    cbind(
    rbind(MakeTable(result1.muslim    [1:28,]),
          format(round(result1.muslim[39,1],2), nsmall=2)),
    rbind(MakeTable(result1.non.muslim[1:28,]),
          round(result1.non.muslim[40,1],2)),
    rbind(MakeTable(result2.muslim    [1:28,]),
          round(result2.muslim[39,1],2)),
    rbind(MakeTable(result2.non.muslim[1:28,]),
          round(result2.non.muslim[40,1],2)),
    rbind(MakeTable(result3.muslim    [1:28,]),
          round(result3.muslim[39,1],2)),
    rbind(MakeTable(result3.non.muslim[1:28,]),
          round(result3.non.muslim[40,1],2))
    )

@
<<Print Table 4, results=tex, echo=FALSE>>=


rownames(Table.Reg)[nrow(Table.Reg)] <- "Adjusted R-square"
varnames <- rownames(Table.Reg)
rownames(Table.Reg) <- NULL
Table.Reg.dat <- data.frame(varnames, Table.Reg)
print(xtable(Table.Reg.dat[1:28,],align=c("llcccccc"),
     caption=paste("Regression Results for Muslims",
     nrow(Umuslim),"and non-Muslims", nrow(UNmuslim),
     "to be compared with Table 2 in \\citet{Bisin08}. ",
     "Heteroskedasticity corrected (HC1) Standard Errors ",
     "are in parentheses. P-values $< 0.05$ are ",
     "marked with *.",sep=" "),
     label="ta:regressions"),
     table.placement = "p",
     floating.environment = "sidewaystable",
     caption.placement="top",
     include.rownames=FALSE,
     include.colnames=FALSE,
     add.to.row=list(pos=list(0),
     command=paste(
     "&\\multicolumn{2}{c}{Importance of}",
     "&\\multicolumn{2}{c}{Inter Ethnic }",
     "&\\multicolumn{2}{c}{Ethnic Composition}\\\\",
     "&\\multicolumn{2}{c}{Religion}",
     "&\\multicolumn{2}{c}{Marriage}",
     "&\\multicolumn{2}{c}{of Schools}",
     "\\\\\\cline{2-7}\\\\",
     "&\\multicolumn{1}{c}{Muslims}",
     "&\\multicolumn{1}{c}{non--Muslims}",
     "&\\multicolumn{1}{c}{Muslims}",
     "&\\multicolumn{1}{c}{non--Muslims}",
     "&\\multicolumn{1}{c}{Muslims}",
     "&\\multicolumn{1}{c}{non--Muslims}\\\\\\hline",
     sep=""))
     ,hline.after=c(-1,0,NULL))
# Table 5: Continuation of Table 4
Footer <- paste("\\hline",
      "\\multicolumn{7}{l}{\\multirow{3}{20cm}",
      "{\\footnotesize NOTE: All estimated models include ",
      "7 UK-region dummies, and ",
      "the variables \\code{DO.NOT.SPEAK.WITH.OLDER}, ",
      "\\code{DO.NOT.SPEAK.AT.WORK}, and ",
      "\\code{DO.NOT.SPEAK.WITH.FRIENDS}.  It turned out ",
      "that the variable \\code{DO.NOT.SPEAK.WITH.YOUNGER} ",
      "is TRUE for few observations and cannot be ",
      "included in the model.}}\\\\",
      sep="")


Header <- paste(
      "&\\multicolumn{2}{c}{Importance of}",
      "&\\multicolumn{2}{c}{Inter Ethnic }",
      "&\\multicolumn{2}{c}{Ethnic Composition}\\\\",
      "&\\multicolumn{2}{c}{Religion}",
      "&\\multicolumn{2}{c}{Marriage}",
      "&\\multicolumn{2}{c}{of Schools}",
      "\\\\\\cline{2-7}\\\\",
      "&\\multicolumn{1}{c}{Muslims}",
      "&\\multicolumn{1}{c}{non--Muslims}",
      "&\\multicolumn{1}{c}{Muslims}",
      "&\\multicolumn{1}{c}{non--Muslims}",
      "&\\multicolumn{1}{c}{Muslims}",
      "&\\multicolumn{1}{c}{non--Muslims}\\\\\\hline",
      sep="")


print(xtable(Table.Reg.dat[29:57,],align=c("llcccccc"),
     caption=paste("Table 4 continued. Regression Results ",
     "for Muslims", nrow(Umuslim)," and non-Muslims",
     nrow(UNmuslim), "to be compared with Table 2 in ",
     "\\citet{Bisin08}. Heteroskedasticity corrected ",
     "(HC1) Standard Errors are in parentheses. P-values ",
     "$< 0.05$ are marked with *.", sep=" "),
      label="ta:regressions2"),
      table.placement = "p",
      floating.environment = "sidewaystable",
      caption.placement="top",
      include.rownames=FALSE,
      include.colnames=FALSE,
      add.to.row=list(pos=list(0,29),
      command=c(Header,Footer))
      ,hline.after=c(-1,0,NULL))
@

%</paper|techreport>

%<*paper|techreport|present>
\end{document}
%</paper|techreport|present>
Here we create the statistics that are cross--references to
in the abstract and the introduction. It is the relative
number of observations that remains after missing values
are removed and its complement:

<<Cross-references,echo=FALSE,keep.source=FALSE>>=
RelativeSampleSize <-round(100*nrow(U)/nrow(FNSEM),
     digits=0)
BisinRelativeSampleSize <- round(100*5963/nrow(FNSEM),
     digits=0)
write(BisinRelativeSampleSize,file="crossreference0.tex")
write(100-RelativeSampleSize,file="crossreference2.tex")
write(nrow(U.Original),file="crossreference3.tex")
write(nrow(U),file="crossreference4.tex")

@


The following code writes the DOCSTRIP commands to the file
araietal.ins:

<<LaTeX, echo=FALSE>>=
write(paste(
     "\\input docstrip.tex",
     "\\keepsilent",
     "\\askforoverwritefalse",
     "\\nopreamble",
     "\\nopostamble",
     "\\generate{",
     "\\file{araietal_paper.tex}{",
     "     \\from{araietal_source.tex}{paper}}",
     "\\file{araietal_techreport.tex}{",
     "     \\from{araietal_source.tex}{techreport}}",
     "\\file{araietal_present.tex}{",
     "     \\from{araietal_source.tex}{present}}",
     "}",
     "\\endbatchfile",
     "\n",sep="\n"),file="araietal.ins")
@

<<BibTex, echo=FALSE>>=
write(paste(
"@Manual{FNSEM,",
"author       = {FNSEM},",
"title        = {P1312 Fourth National Study of Etnic Minorities. {P}roject instructions},",
"note         = {Project 3685},",
"organization = {Social and Community Planning Research}, ",
"address      = {London},",
"year         = {1993},",
"url          = {http://www.data-archive.ac.uk/doc/3685/mrdoc/pdf/a3685uab.pdf}",
"}",
"\n",
"@Manual{FNSEM93b,",
"author       = {FNSEM},",
"title        = {UK Data Archive Data Dictionary}, ",
"note         = {An RTF file called \\url{UKDA-3685-tab/mrdoc/allissue/3685_UKDA_Data_Dictionary.rtf} available when the FNSEM data set is downloaded from the UK Data Archive}, ",
"organization = {Social andCommunity Planning Research}, ",
"address      = {London},",
"year         = {1993}",
"}",
"\n",
"@Article{ Bisin08,",
"author       = {Alberto Bisin and Eleonora Patacchini and Thierry Verdier and Yves Zenou},",
"title        = {Are Muslim Immigrants Different in Terms of Cultural Integration?},",
"journal      = {Journal of the European Economic Association},",
"pages        = {445--456},",
"volume       = {6},",
"year         = {2008}",
"}",
"\n",
"@Manual{Rcore,",
"title        = {R: A Language and Environment for Statistical Computing},",
"author       = {{R Development Core Team}},",
"organization = {R Foundation for Statistical Computing},",
"address      = {Vienna, Austria},",
"year         = {2008},",
"note         = {{ISBN} 3-900051-07-0},",
"url          = {http://www.R-project.org}",
"}",
"\n",
"@Unpublished{Koenker07,",
"author      = {Koenker, Roger and Zeileis, Achim},",
"year        = {2007},",
"title       = {Reproducible Econometric Research (A Critical Review of the State of the Art)},",
"note        = {Report 60, Department of Statistics and Mathematics, Wirtschaftsuniversität Wien, Research Report Series},",
"url         = {http:epub.wu-wien.ac.at/dyn/virlib/wp/eng/mediate/epub-wu-01_c75.pdf?ID=epub-wu-01_c75}",
"}",
"\n",
"@Unpublished{Araietal08a,",
"author      = {Arai, Mahmood and Karlsson, Jonas and Lundholm, Michael},",
"year        = {2008},",
"title       = {On Fragile Grounds: A replication of \\emph{Are Muslim immigrants different in terms of cultural
integration?}},",
"note        = {Unpublished manuscript}",
"}",
"\n",
"@Unpublished{Araietal08b,",
"author      = {Arai, Mahmood and Karlsson, Jonas and Lundholm, Michael},",
"year        = {2008},",
"title       = {On Fragile Grounds: {A} replication of \\emph{Are Muslim immigrants different in terms of cultural integration? {T}echnical documentation}},",
"note        = {Unpublished manuscript}",
"}",
"\n",
"@Unpublished{Leisch02,",
"author      = {Friedrich Leisch},",
"title       = {Sweave User Manual},",
"year        = {2002},",
"url         = {http://www.ci.tuwien.ac.at/~leisch/Sweave/}",
"}",
"\n",
"@Manual{Berthoud,",
"title       = {Fourth National Survey of Ethnic Minorities, 1993-1994 [computer file]},",
"author      = {Berthoud, R.G. and Modood, T. and Smith, P. and Prior, G.},",
"publisher   = {UK Data Archive [distributor]},",
"address     = {Colchester, Essex: UK},",
"year        = {1997},",
"note        = {SN: 3685},",
"url         = {http://www.data-archive.ac.uk/doc/3685%5Cmrdoc%5CUKDA%5CUKDA_Study_3685_Information.htm}",
"}",
sep="\n"),file="araietal.bib")
@

<<Extracting R code,results=hide>>=
Stangle("araietal_source.Rnw")
@

<<Script to make LaTeX files,echo=FALSE,silent=TRUE,results=hide>>=
write(paste(
"system(\"latex araietal.ins\")",
"system(\"pdflatex araietal_techreport.tex\")",
"system(\"bibtex araietal_techreport\")",
"system(\"pdflatex araietal_techreport.tex\")",
"system(\"pdflatex araietal_techreport.tex\")",
"system(\"latex araietal_paper.tex\")",
"system(\"bibtex araietal_paper\")",
"system(\"latex araietal_paper.tex\")",
"system(\"latex araietal_paper.tex\")",
"system(\"dvips araietal_paper.dvi\")",
"system(\"ps2pdf araietal_paper.ps\")",
"system(\"latex araietal_present.tex\")",
"system(\"dvips -o araietal_present1.ps -pp 1:14 araietal_present.dvi\")",
"system(\"dvips -o araietal_present2.ps -pp 7:11 -x 1095 araietal_paper.dvi\")",
"system(\"dvips -o araietal_present3.ps -p 20 araietal_present.dvi\")",
"system(\"gs -q -r2400 -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -sOutputFile=araietal_present.pdf -f  araietal_present1.ps araietal_present2.ps araietal_present3.ps\")",
"system(\"pdflatex araietal_paper.tex\")",
sep="\n"),file="script.R")
@