[Rsiena-commits] r73 - in pkg/RSienaTest: R doc inst/doc
noreply at r-forge.r-project.org
noreply at r-forge.r-project.org
Tue Mar 30 19:58:58 CEST 2010
Author: ripleyrm
Date: 2010-03-30 19:58:55 +0200 (Tue, 30 Mar 2010)
New Revision: 73
Modified:
pkg/RSienaTest/R/print01Report.r
pkg/RSienaTest/doc/RSienaDeveloper.tex
pkg/RSienaTest/doc/s_man400.tex
pkg/RSienaTest/inst/doc/s_man400.pdf
Log:
Latest versions of documentation and change to text on report
Modified: pkg/RSienaTest/R/print01Report.r
===================================================================
--- pkg/RSienaTest/R/print01Report.r 2010-03-30 17:57:24 UTC (rev 72)
+++ pkg/RSienaTest/R/print01Report.r 2010-03-30 17:58:55 UTC (rev 73)
@@ -377,8 +377,8 @@
Report(c("Missing values in this actor variable are",
"imputed",
"by the mode per observation.\n"), outf)
- Report(c("But if there is a previous nonmissing",
- "value,",
+ Report(c("But if there is a previous (or later)",
+ "nonmissing value,",
"this is used as the imputed value.\n"), outf)
Report("Modal values:\nObservation ", outf)
Report(c(format(1:x$observations+periodFromStart,
Modified: pkg/RSienaTest/doc/RSienaDeveloper.tex
===================================================================
--- pkg/RSienaTest/doc/RSienaDeveloper.tex 2010-03-30 17:57:24 UTC (rev 72)
+++ pkg/RSienaTest/doc/RSienaDeveloper.tex 2010-03-30 17:58:55 UTC (rev 73)
@@ -82,6 +82,7 @@
\subsection{R}
\begin{enumerate}
\item Use \verb|<-| rather than \verb|=| for assignment.
+\item Use \sfn{TRUE} and \sfn{FALSE} not \sfn{T} and \sfn{F}.
\item Avoid unnecessary \sfn{for} loops. They are acceptable where a complicated
\sfn{apply} statement would be necessary, but not where a simple vectorised
command could be used.
Modified: pkg/RSienaTest/doc/s_man400.tex
===================================================================
--- pkg/RSienaTest/doc/s_man400.tex 2010-03-30 17:57:24 UTC (rev 72)
+++ pkg/RSienaTest/doc/s_man400.tex 2010-03-30 17:58:55 UTC (rev 73)
@@ -5,7 +5,11 @@
%Required files: pdfscreen.sty, pdfscreen.cfg, ilcampo_bg.jpg, ilcampo.jpg
%\usepackage{times}
\usepackage{natbib}
+\usepackage{rotating}
+\usepackage{longtable, lscape}
+\usepackage{threeparttablex}
\usepackage{amsmath}
+ \usepackage[top=2.5cm, bottom=2.5cm, left=2cm , right=1.8cm]{geometry}
%\usepackage[bookmarksopen=false]{hyperref}
% in the newer version of pdfscreen, hyperref can be loaded first with its own parameters
\usepackage[pdftex,dvipsnames]{color}
@@ -20,12 +24,12 @@
\usepackage{pictexwd}
-\usepackage{supertabular}
-\usepackage{tabls}
+%\usepackage{supertabular}
+%\usepackage{tabls}
\usepackage{enumitem}
-
+\setlength{\bibsep}{0.01in}
\begin{screen}
\margins{.65in}{.65in}{.65in}{.65in}
\screensize{6.25in}{8in}
@@ -50,6 +54,8 @@
%\renewcommand\textfraction{0}
%\def\pdfscreen{\texttt{\small\color{section1}pdfscreen}\xspace}
+%\def\bibsection{\section{\refname}}
+\renewcommand\bibsection{\section{\refname}}
\newcommand{\opmerking}[1]{\par \fbox{\Large #1} \par}
%\newcommand{\opmerking}[1]{}
@@ -94,7 +100,7 @@
\newcommand{\ga}[1]{$\gamma_{#1}$}
\newcommand{\beq}{\begin{equation}}
\newcommand{\eeq}{\end{equation}}
-\renewcommand{\bibitem}[1]{\bigskip \par \noindent \hspace{-4pt}}
+%\renewcommand{\bibitem}[1]{\bigskip \par \noindent \hspace{-4pt}}
\makeatletter
\newenvironment{indentation}[2]
{\par \setlength{\leftmargin}{#1} \setlength{\rightmargin}{#2}
@@ -102,7 +108,7 @@
\advance\@totalleftmargin\leftmargin \@setpar{{\@@par}}%
\parshape 1 \@totalleftmargin \linewidth \ignorespaces}{\par}
\makeatother
-\renewcommand{\bibitem}[1]{\par \noindent \hskip-\parindent}
+%\renewcommand{\bibitem}[1]{\par \noindent \hskip-\parindent}
\newcommand{\separationb}{\\[0.5ex]\hline\rule{0pt}{2ex}}
@@ -172,8 +178,8 @@
\noindent \SI (for {\sf Simulation Investigation for Empirical
Network Analysis}) is a computer program that carries out the
statistical estimation of models for the evolution of social
-networks according to the dynamic actor-oriented model of Snijders
-(2001, 2005) and Snijders, Steglich, and Schweinberger (2007).
+networks according to the dynamic actor-oriented model of \citet{Snijders01,
+Snijders05} and \citet{SnijdersEA07}.
This is the manual for \SI version 4, which is a contributed package to
the statistical system \Rn.
The manual is based on the earlier manual for \SI version 3,
@@ -213,28 +219,26 @@
\footnote{This program was first presented at the
International Conference for Computer Simulation and the Social
Sciences, Cortona (Italy), September 1997, which originally was
-scheduled to be held in Siena. See Snijders \& van Duijn (1997).}
+scheduled to be held in Siena. See \citet{SnijdersDuijn97}.}
\end{print}
\begin{screen}
\footnote{This program was first presented
at the International Conference for Computer Simulation and the
Social Sciences, Cortona (Italy), September 1997, which originally
-was scheduled to be held in Siena. See Snijders \& van Duijn (1997).
+was scheduled to be held in Siena. See \citet{SnijdersDuijn97} .
The background picture in this manual is the Palazzo Pubblico with
the Torre del Mangia in Siena.}
\end{screen}
$\!\!\!$, shorthand for {\sf Simulation Investigation for Empirical
Network Analysis}, is a computer program that carries out the
statistical estimation of models for repeated measures of social
-networks according to the dynamic actor-oriented model of Snijders
-and van Duijn (1997), Snijders (2001), and
-Snijders, Steglich, and Schweinberger (2007); also see
-Steglich, Snijders, and Pearson (2010).
-A tutorial for these models is in Snijders, van de Bunt, and Steglich (2010).
+networks according to the dynamic actor-oriented model of \citet{SnijdersDuijn97}, \citet{Snijders01}, and
+\citet*{SnijdersEA07}; also see
+\citet*{SteglichEA10}.
+A tutorial for these models is in \citet*{SnijdersEA10b}.
Some examples are
-presented, e.g., in van de Bunt (1999); van de Bunt, van Duijn, and
-Snijders (1999); and van Duijn, Zeggelink, Stokman, and Wasseur (2003);
-and Steglich, Snijders, and West (2006).
+presented, e.g., in \citet*{vanBunt99, vanBuntEA99} and \citet*{vanDuijnEA03};
+and \citet*{SteglichEA06}.
A website for \SI is maintained at \url{http://www.stats.ox.ac.uk/~snijders/siena/}~.
At this website
@@ -492,7 +496,7 @@
window.)
\subsection{Entering Data.}
-
+\label{thegui}
There are two ways to enter the data.
\begin{enumerate}
\item Enter each of your data files using \sfn{Add}.\\
@@ -507,7 +511,7 @@
by the \sfn{ siena01Gui()} when you request
\sfn{Save to file}.
\item If you wish to remove files, use the \sfn{Remove} option rather than
- blanking out the entries.
+ blanking out the entries.
\end{enumerate}
Once you have done this, check that the \sfn{Format},
\sfn{Period}, \sfn{Type}, etc., are correct, and enter any
@@ -526,6 +530,7 @@
\subsection{Running the Estimation Program}
+\label{estgui}
\begin{enumerate}
\item Click \sfn{Apply}: you will be prompted to save your work. Then you should
see the \sfn{Model Options} screen shown in \hyperlink{options}{Figure
@@ -607,9 +612,9 @@
\item[\sfn{changing dyadic covariate}]
\item[\sfn{exogenous event}] (for changing composition of the actor set)
\end{description}
-\item[\sfn{Selected}] Yes or No. Files with \sfn{Yes} \emph{or blank} will be
+\item[\sfn{Selected}] Yes or No. Files with \sfn{Yes} \emph{or blank} will be
included in the model. Use this field to remove any networks or behavior
-variables that are not required in the model.
+variables that are not required in the model.
\item[\sfn{Missing Values}] Enter any values which indicate missingness, with
spaces between different entries.
\item[\sfn{Nonzero Codes}] Enter any values which indicate ties, with spaces
@@ -649,7 +654,7 @@
network in the sense that some actors are not part of the network during
all the observations.
This will trigger treatment of such change of composition
-according to Huisman and Snijders (2003).
+according to \citet{HuismanSnijders03}.
This file must have one row for each node.
Each row should be
consist of a set of pairs of numbers which indicate the periods
@@ -1337,17 +1342,16 @@
\smallskip
-A basic reference for the network dynamics model is Snijders (2001)
-or Snijders (2005).
+A basic reference for the network dynamics model is \citet{Snijders01}
+or \citet{Snijders05}.
Basic references for the model of network-behavior co-evolution
-are Snijders, Steglich, and Schweinberger (2007)
-and Steglich, Snijders, and Pearson (2010).
+are \citet*{SnijdersEA07} and \citet*{SteglichEA10}.
-More specific references are Schweinberger (2005) for the score-type goodness of
-fit tests and Schweinberger and Snijders (2007) for the calculation of standard
-errors of the Method of Moments estimators .
+More specific references are \citet{Schweinberger10} for the score-type goodness
+of fit tests and \citet{SchweinbergerSnijders07a} for the calculation of
+standard errors of the Method of Moments estimators .
-A tutorial is Snijders, van de Bunt, and Steglich (2010).
+A tutorial is \citet*{SnijdersEA10b}.
\subsection{Getting help with problems}
@@ -1784,8 +1788,7 @@
\SI also allows dependent action variables,
also called dependent behavior variables. This can be used in studies
of the co-evolution of networks and behavior, as described
-in Snijders, Steglich, and Schweinberger (2007) and Steglich, Snijders,
-and Pearson (2010).
+in \citet*{SnijdersEA07} and \citet*{SteglichEA10}.
These action variables represent the actors' behavior, attitudes, beliefs, etc.
The difference between dependent action variables and changing actor
covariates is that the latter change exogenously, i.e., according
@@ -1832,7 +1835,7 @@
In the current implementation of \si, missing data are treated in
a simple way, trying to minimize their influence on the estimation
results.
-This method is further explained in Huisman and Steglich (2008),
+This method is further explained in \citet{HuismanSteglich08},
where comparisons are also made with other ways of dealings with the missing
information.
@@ -1856,7 +1859,7 @@
if there is an earlier observed value of this variable then
the last observed value is used to impute the current
value (the `last observation carry forward' option,
-cf.\ Lepkowski, 1989); if there is no earlier observed
+cf.\ \citet{Lepkowski89}; if there is no earlier observed
value, the value 0 is imputed.
For the dependent behavior variables the same principle
is used: if there is a previous observation of the same variable
@@ -1889,8 +1892,8 @@
\SI can also be used to analyze networks of which the composition
changes over time, because actors join or leave the network
between the observations.
-This can be done in two ways: using the method of Huisman and Snijders
-(2003), or using structural zeros.
+This can be done in two ways: using the method of \citet{HuismanSnijders03},
+or using structural zeros.
(For the maximum likelihood estimation option, the Huisman-Snijders method
is not implemented, and only the structural zeros method can be used.)
Structural zeros can specified for all elements of the tie variables
@@ -1990,8 +1993,7 @@
% the file \textsf{{\em pname}.eff}.
For the longitudinal case, three types of
-effects are distinguished (see Snijders, 2001; Snijders,
-van de Bunt, and Steglich, 2010):
+effects are distinguished \citep*[see][]{Snijders01, SnijdersEA10b}:
\begin{itemize}
\item {\em rate function effects}\\
@@ -2012,8 +2014,7 @@
%See XXXXXXX.
\item {\em evaluation function effects}\\
The evaluation function\footnote{The evaluation function was called
-\emph{objective function} in Snijders,
-2001.} models the network actors' satisfaction with their local
+\emph{objective function} in \citet{Snijders01}} models the network actors' satisfaction with their local
network neighborhood configuration. It is assumed that actors
change their scores on the dependent variable such that they
improve their total satisfaction -- with a random element
@@ -2036,7 +2037,7 @@
to account more precisely for the distribution of the behavior.
\item {\em endowment function effects}\\
The endowment function\footnote{The endowment function is similar to the {\it gratification
-function} in Snijders, 2001.} is an extension of the evaluation
+function} in \citet{Snijders01}} is an extension of the evaluation
function that allows to distinguish between new and old network
ties (when evaluating possible network changes) and between
increasing or decreasing behavioral scores (when evaluating
@@ -2273,7 +2274,7 @@
\item \begin{minipage}[t]{.6\textwidth}
Transitivity in two-mode networks is expressed in the first
- place by the number of \emph{four-cycles} (Robins and Alexander, 2005).
+ place by the number of \emph{four-cycles} \citep{RobinsAlexander04}.
This reflects the extent to which actors who make one choice in common
also make other choices in common.
\vfill
@@ -2575,7 +2576,7 @@
For models with one or more dependent behavior variables, i.e.,
models for the co-evolution of networks and behavior,
the most important effects for the behavior dynamics are the following;
-see Steglich, Snijders, and Pearson (2010).
+see \citet*{SteglichEA10}.
In these descriptions, with the `alters' of an actor
we refer to the other actors to whom
the focal actor has an outgoing tie.
@@ -2680,9 +2681,9 @@
\iffalse
For directed networks, the Model Type distinguishes between
-the model of Snijders (2001) (Model Type 1), that of Snijders
-(2003) (Model Type 2),
-and the tie-based model described in Snijders (2006) (Model Type 3).
+the model of \citet{Snijders01} (Model Type 1),
+that of \citet{Snijders03} (Model Type 2),
+and the tie-based model described in \citet{Snijders06} (Model Type 3).
Model Type 1 is the default model and is
described in the basic publications on Stochastic Actor-Oriented
Models for network dynamics.
@@ -2710,8 +2711,7 @@
connected to the parameters for the structural dynamics. The use of
such an approach in statistical modeling minimizes the influence of
the observed degrees on the conclusions about the structural aspects
-of the network dynamics. This is further explained in Snijders
-(2003).
+of the network dynamics. This is further explained in \citet{Snijders03}.
For Model Type 2, in the rate function, effects connected to these
functions $\xi$ and $\nu$ are included. On the other hand, effects
@@ -2725,7 +2725,7 @@
given. For Model Type 1, this comparison is given by adding 10 to the
Model Code in the advanced options. (For \LaTeX\ users: the log
file contains code that can be used to make a graph of the type
-given in Snijders, 2003).
+given in \citet{Snijders03}.
For using Model Type 2, it is advised to first estimate some model
according to Model Type 1 (this may be a simple model containing a
@@ -2749,7 +2749,7 @@
\item the factorial out-degree effect
\item the logarithmic out-degree effect.
\end{enumerate}
-These are the effects defined in formula (18) of Snijders (2003b)
+These are the effects defined in formula (18) of \citet{Snijders03}
and indicated with the parameters $\alpha_1$, $\alpha_2$,
and $\alpha_3$, respectively.
The user has to see from the estimation results which, or which two,
@@ -2761,8 +2761,8 @@
\iffalse
In addition these types, there is
-Model Type 6 which implements the reciprocity model of Wasserman (1979)
-and Leenders (1995) (also see Snijders, 1999, 2005) ---
+Model Type 6 which implements the reciprocity model of \citet{Wasserman79}
+and \citet{Leenders95} \citep[also see][]{Snijders99, Snijders05} ---
provided that no other effects are chosen than
the outdegree effect, the reciprocity effect and perhaps
the reciprocity endowment effect,
@@ -2785,13 +2785,13 @@
\label{S_modeltype_nd}
Non-directed networks are an undocumented option (there currently
-only is the presentation Snijders 2007), and therefore
+only is the presentation \citet{Snijders07}, and therefore
mentioned here reluctantly for those users who want to use
this option anyway.
\SI detects automatically when the networks all are non-directed, and then employs a model for this
special case. For non-directed networks, the Model Type has seven possible values,
-as described in Snijders (2007).
+as described in \citet{Snijders07}.
\begin{enumerate}
\item Forcing model: \\
@@ -3280,7 +3280,7 @@
then actor heterogeneity can be taken into account by including covariates in the model.
If not all relevant actor heterogeneity is observed,
then more complex models are required, such as random effects models.
-Random effects models (see Schweinberger and Snijders, 2007a) allow to take
+Random effects models \citep[see][]{SchweinbergerSnijders07b} allow to take
unobserved actor heterogeneity into account
by assuming that the network (and behavior) evolution is affected by unobserved
outcomes of actor-dependent random variables (random effects),
@@ -3331,11 +3331,10 @@
approximation algorithm.
%Three estimation procedures are implemented:
Only one estimation procedure is currently implemented:
-the Method of Moments (MoM) (Snijders, 2001; Snijders, Steglich,
-and Schweinberger, 2007);
-% the Method of Maximum Likelihood (ML) (Snijders, Koskinen and Schweinberger, 2007);
-% and a Bayesian method (Koskinen, 2005; Koskinen and Snijders, 2007;
-% Schweinberger and Snijders, 2007).
+the Method of Moments (MoM) \citep*{Snijders01, SnijdersEA07};
+% the Method of Maximum Likelihood (ML) \citep{SnijdersEA10};
+% and a Bayesian method \citep{Koskinen04, KoskinenSnijders07,
+% SchweinbergerSnijders07c).
% For non-constant rate functions, currently only
% MoM estimation is available.
% The Method of Moments is the default;
@@ -3396,8 +3395,9 @@
\subsection{\label{algorithm}Algorithm} %MS
The estimation algorithm
-is an implementation of the Robbins-Monro (1951) algorithm,
-described in Snijders (2001, 2002), and
+is an implementation of the Robbins-Monro \citeyearpar{RobbinsMonro51}
+ algorithm,
+described in \citet{Snijders01, Snijders02}, and
has %for both the MoM and ML method
three phases:
\begin{enumerate}
@@ -3485,8 +3485,8 @@
The primary information in the output of the estimation process
consists of the following three parts. Results are presented here
-which correspond to Table 2, column ``$t_1$, $t_3$" of Snijders
-(2001). The results were obtained in an independent repetition of
+which correspond to Table 2, column ``$t_1$, $t_3$" of \citet{Snijders01}.
+The results were obtained in an independent repetition of
the estimation for this data set and this model specification;
since the repetition was independent, the results are slightly
different, illustrating the stochastic nature of the estimation
@@ -3723,13 +3723,12 @@
\label{S_Bayes}
\SI can estimate models by three estimation methods: the (unconditional or conditional)
-Method of Moments (`MoM', the default; Snijders, 2001; Snijders, Steglich, and Schweinberger 2007),
-the Maximum Likelihood method (`ML', see Snijders, Koskinen, and Schweinberger, 2010),
-and Bayesian methods (see
-Koskinen, 2005; Koskinen and Snijders, 2007; Schweinberger and Snijders, 2007b).
+Method of Moments \citep*[`MoM', the default;][]{Snijders01; SnijdersEA07},
+the Maximum Likelihood method \citep[`ML', see][]{SnijdersEA10},
+and Bayesian methods
+\citep[see][]{Koskinen04, KoskinenSnijders07, SchweinbergerSnijders07c}.
The maximum likelihood and Bayesian procedures are not yet
implemented in RSiena.
-
In nice situations (relatively small and large network data sets,
and large network and behavior data sets),
the three methods tend to agree
@@ -3791,7 +3790,7 @@
The {\tt R} functions input files generated by {\tt Siena} and output,
among other things,
trace plots and MCMC lag $1, \dots, 100$ autocorrelations of sampled entities
-(see Schweinberger and Snijders, 2007a,b),
+\citep[see][]{SchweinbergerSnijders07b, SchweinbergerSnijders07c},
and,
in the Bayesian case,
in addition $95\%$ posterior intervals, histograms, and Gaussian kernel density
@@ -3866,8 +3865,8 @@
\hyperlink{T_convergence}{lack of convergence of the algorithm}.
(This type of problem also occurs in maximum likelihood estimation
for logistic regression and certain other generalized linear
-models; \label{LargeFix} see Geyer and Thompson (1992, Section
-1.6), Albert and Anderson (1984), Hauck and Donner (1978).)
+models; \label{LargeFix} see \citet[section 1.6]{GeyerThompson92},
+\citet{AlbertAnderson84, HauckDonner77}.)
In such cases this effect
should be fixed to some large value and not left free to be
estimated. This can be specified in the model specification
@@ -4004,21 +4003,20 @@
\item[(1)] score function method 1 (default),
\item[(2)] score function method 2 (not currently implemented).
\end{itemize}
-Schweinberger and Snijders (2006) point out that the finite differences method is associated with a bias-variance dilemma,
-and proposed the unbiased and consistent score function methods.
-These methods demand less computation time than method (0).
-\iffalse
-Method 1 estimates the derivatives per observation period separately
-by the simulated sample covariance of the complete data score
-function and the generated statistics; this is then added over
-the observation periods.
-Especially for more than 2 observations, method 1 has a much
-smaller standard error of the estimated standard errors than the other methods.
-\fi
-It is recommended to use at least 1000 iterations (default) in phase 3.
-For published results, it is recommended to have 2000 or 4000 iterations
-in phase 3.
+\citet{SchweinbergerSnijders07a} point out that the finite differences method is
+associated with a bias-variance dilemma, and proposed the unbiased and
+consistent score function methods. These methods demand less computation time
+than method (0).
+\iffalse
+Method 1 estimates the derivatives per observation
+period separately by the simulated sample covariance of the complete data score
+function and the generated statistics; this is then added over the observation
+periods. Especially for more than 2 observations, method 1 has a much smaller
+standard error of the estimated standard errors than the other methods. \fi It
+is recommended to use at least 1000 iterations (default) in phase 3. For
+published results, it is recommended to have 2000 or 4000 iterations in phase 3.
+
\begin{print}
\newpage
\end{print}
@@ -4041,8 +4039,7 @@
In the maximum likelihood estimation method
it is possible to request likelihood ratio tests.
The log likelihood ratio is computed
-by bridge sampling (Gelman and Meng, 1998;
-Handcock and Hunter, 2006).
+by bridge sampling \citep{GelmanMeng98, HandcockHunter06}.
This can be requested (a bit deviously) by the number of runs in phase 3
(defined in the \hyperlink{T_S_options}{specification options}):
\begin{enumerate}
@@ -4071,7 +4068,7 @@
in \SI (see Schweinberger, 2005).
\iffalse
For the ML estimation method,
-following the same steps produces the Rao (1947) efficient score test.
+following the same steps produces the \citet{Rao47} efficient score test.
\fi
Most goodness-of-fit tests will have the following form: some model
@@ -4093,7 +4090,7 @@
myeff[10, 'test'] <- TRUE
myeff[10, 'initialValue'] <- ((value to be used for test))
## or, more easily
-myeff <- setEffect(myeff, recip, fix=TRUE, test=TRUE,
+myeff <- setEffect(myeff, recip, fix=TRUE, test=TRUE,
initialValue=(value to be used for test))
\end{verbatim}
@@ -4330,7 +4327,7 @@
For simulating networks and behavior, the output includes
the autocorrelation statistics known as Moran's $I$ and Geary's $c$.
-For formulae and interpretation see, e.g., Ripley (1981, 98--99).
+For formulae and interpretation see, e.g., \citet[98--99]{Ripley81}.
These measure the extent to which the value of the variable
in question is similar between tied actors.
This similarity is expressed by relatively high values for Moran's $I$
@@ -4411,7 +4408,7 @@
% In the longitudinal case, the meaning of this code is as follows.\\
% Model Codes 10 or more give extra output for evaluating the fit of
% the out-degree distribution and for the explained variation
-% (Snijders, 2004);\\
+% \citet{Snijders04};\\
% the integer Model Code in the unit position (i.e.,
% Model Code itself if it is less than 10, and Model Code - 10 if the code is more than 10)
% defines the Model Type defined in Section~\ref{S_modeltype}.\\[0.5ex]
@@ -4439,7 +4436,7 @@
% MCMC algorithm.
\item The initial gain value, which is the step size in the starting
steps of the Robbins-Monro procedure, indicated in
- Snijders (2001) by $a_1\,$.
+ \citet{Snijders01} by $a_1\,$.
\item The choice between standard initial values (suitable
estimates for the density and reciprocity parameters and zero
values for all other parameters) or
@@ -4461,7 +4458,7 @@
%(this is the method used in
%\SI versions 1 and 2, which has a bias);
1 is the more efficient and unbiased
- method proposed by Schweinberger and Snijders (2007);
+ method proposed by \citet{SchweinbergerSnijders07a};
this is the preferred method. See Section~\ref{S_se}.
\end{enumerate}
@@ -4486,8 +4483,8 @@
For getting a first acquaintance with the model, one may use the
data set collected by Gerhard van de Bunt, discussed extensively in
-van de Bunt (1999), van de Bunt, van Duijn, and Snijders (1999),
-and used as example also in Snijders (2001) and Snijders (2005).
+\citet*{vanBunt99, vanBuntEA99},
+and used as example also in \citet{Snijders01} and \citet{Snijders05}.
The data files are provided with the program
and at the \SI website. The digraph data files
used are the two networks {\sf vrnd32t2.dat}, {\sf vrnd32t4.dat}.
@@ -4961,10 +4958,11 @@
estimated for several data sets.
\textsf{Siena08} combines
the estimates in a meta-analysis or multilevel analysis
- according to the methods of Snijders and Baerveldt (2003),
+ according to the methods of \citet{SnijdersBaerveldt03},
and according to a Fisher-type combination of one-sided $p$-values.
- This combination method of Fisher (1932) is described in Hedges and Olkin (1985)
- and (briefly) in Snijders and Bosker (1999, Chapter 3).
+ This combination method of \citet{Fisher32} is described in
+\citet{HedgesOlkin85}
+ and (briefly) in \citet[Chapter 3]{SnijdersBosker99}).
Some more information is at the \SI website.
For \SI version 4 the program \textsf{Siena08.exe} still must be
@@ -5047,7 +5045,7 @@
this is important for two reasons:<br>
1. it allows to see easily how many positive and negative individually significant
parameter values are contained in the combined data set;<br>
- 2. an assumption of the Snijders-Baerveldt (2003)
+ 2. an assumption of the Snijders-Baerveldt \citeyearpar{SnijdersBaerveldt03}
method for meta-analysis is that standard errors and true parameter values
are uncorrelated; this can be visually checked from this plot.
<li>An extra method for combining the various classes, which does not make this assumption.<br>
@@ -5148,8 +5146,8 @@
Here, the mathematical formulae for the definition of the effects
-are given. In Snijders (2001, 2005) and Steglich, Snijders and Pearson,
-(2010), further background to these formulae can be found.
+are given. In \citet{Snijders01,Snijders05} and \citet*{SteglichEA10},
+further background to these formulae can be found.
The effects are grouped into effects for modelling network
evolution and effects for modelling behavioral evolution (i.e.,
the dynamics of dependent actor variables). Within each group of
@@ -5197,7 +5195,8 @@
this is indicated below by ``endowment effect only likelihood-based''.
(It may be noted that the network evaluation function was called objective function,
-and the endowment function was called gratification function, in Snijders, 2001.)
+and the endowment function was called gratification function, in
+\citet{Snijders01}.)
\subsubsection{Network evaluation function}
\label{S_f}
@@ -5938,10 +5937,9 @@
The potential effects $s^{\rm net}_{ik}(x) $ in this function, and their
formulae, are the same as in the evaluation function;
except that not all are available, as indicated in the preceding subsection.
-For further explication, consult Snijders (2001, 2005;
-here, the `gratification function' is used rather than the endowment function),
-Snijders, Steglich, and Schweinberger (2007), and Steglich, Snijders
-and Pearson (2010).
+For further explication, consult \citet{Snijders01, Snijders05};
+(here, the `gratification function' is used rather than the endowment function),
+\citet*{SnijdersEA07}, and \citet*{SteglichEA10}.
\begin{screen}
\newpage
@@ -5953,7 +5951,7 @@
Type) as a product \[ \lambda^{\rm net}_i(\rho, \alpha, x, m) =
\lambda^{\rm net}_{i1} \lambda^{\rm net}_{i2} \lambda^{\rm net}_{i3}
\] of factors depending, respectively, on period $m$, actor
-covariates, and actor position (see Snijders, 2001, p.\ 383). The
+covariates, and actor position \citep[see][p.\ 383]{Snijders01}. The
corresponding factors in the rate function are the following:
\begin{enumerate}
\item The dependence on the period can be represented by a simple factor
@@ -5977,7 +5975,7 @@
on the out-degree is represented by
\[ \lambda^{\rm net}_{i3} = \frac{x_{i+}}{n-1} \exp(\alpha_1) \+
\left(1 - \frac{x_{i+}}{n-1}\right) \exp(- \alpha_1). \]
- This formula is motivated in Snijders and Van Duijn (1997).
+ This formula is motivated in \citet{SnijdersDuijn97}.
This defines a linear function of the out-degree,
parametrized in such a way that it is necessarily positive.\\
For a general dependence on the out-degree, in-degree, and number
@@ -6011,7 +6009,7 @@
\subsubsection{Network rate function for Model Type 2}
For Model Type 2 (see Section~\ref{S_modeltype}), the network rate
-function is defined according to Snijders (2003) by
+function is defined according to \citet{Snijders03} by
\begin{eqnarray*}
\rho_m\, \lambda_{i+}(s) & = & \rho_m\,\frac{\nu(s)\, \xi(s)}{1 \,+\, \xi(s)}\, , \\
\rho_m\, \lambda_{i-}(s) & = & \rho_m\, \frac{\nu(s-1)}{1 \,+\, \xi(s-1)} \ ,
@@ -6023,8 +6021,7 @@
period.
Function $\xi$ (\emph{xi}) is called the distributional tendency
-function and is represented according to Snijders (2003, formula
-(17)) by
+function and is represented according to \citet[formula (17)]{Snijders03} by
\[ \xi(s) \,=\, \exp\left(\alpha_1 \,-\, \alpha_2 \log(s+1) - \frac{\alpha_3}{s+1}\right) \ . \]
where the names given in \SI are
\begin{itemize}
@@ -6033,7 +6030,7 @@
\item $\alpha_3$ : factorial out-degree effect.
\end{itemize}
The reasons for these names and interpretation of the effects
-can be found in Snijders (2003).
+can be found in \citet{Snijders03}.
To the exponent also effects of actor covariates can be added.
The so-called volatility function $\nu$ (\emph{nu}) is defined as
@@ -6237,8 +6234,8 @@
decreasing his behavioral score by one unit (downward steps), not
when upward steps (or no change) are considered. For more details,
consult
-Snijders, Steglich, and Schweinberger (2007) and
-Steglich, Snijders and Pearson (2010).
+\citet*{SnijdersEA07} and
+\citet*{SteglichEA10}.
The statistics reported as \emph{dec.\ beh.} (decrease in behavior)
are the sums of the changes in actor-dependent values
@@ -6270,8 +6267,9 @@
The main `driving force' of the actor-oriented model
is the evaluation function
-(in earlier publications called objective function,
-see Snijders, 2001, 2005) given in formula (\ref{f_net}) (for the network) as
+\citep[in earlier publications called objective function,
+see][]{Snijders01, Snijders05} given in formula (\ref{f_net})
+(for the network) as
\[
f^{\rm net}(x) \, = \, \sum_k \beta^{\rm net}_k \, s^{\rm net}_{ik}(x) \ .
\]
@@ -6284,7 +6282,7 @@
to the values of the objective function with standard deviations
equal\footnote{More exactly, the value is $\sqrt{\pi^2/6}$,
the standard deviation of the Gumbel
-distribution; see Snijders (2001).} to 1.28.
+distribution; see \citet{Snijders01}.} to 1.28.
An alternative interpretation is that when actor $i$ is making
a `ministep', i.e., a single change in his outgoing ties
@@ -6392,8 +6390,8 @@
This can be concretely carried using the data set {\sf s50}
which is an excerpt of 50 girls in the data set used in
-Pearson and Michell (2000), Pearson and West (2003),
-Steglich et al.\ (2006) and Steglich et al.\ (2007).
+\citet{PearsonMichell00, PearsonWest03,
+SteglichEA06} and \citet{SteglichEA07}.
We refer to any of these papers for a further description of the data.
The friendship network data over 3 waves are in
the files {\sf s50-network1.dat}, {\sf s50-network2.dat},
@@ -6807,12 +6805,488 @@
%\end{print}
%\fi
+%% \documentclass[a4paper,11pt,titlepage]{article}
+% \usepackage{rotating}
+% \usepackage{longtable, lscape}
+% \usepackage[top=2.5cm, bottom=2.5cm, left=2cm , right=1.8cm]{geometry}
+% \author{Paulina Preciado}
+% \title{List of Functions for RSiena}
+
+% \begin{document}
+% \maketitle
+% \tableofcontents
+% \listoftables
+\appendix
+\newpage
+\section{List of Functions in Order of Execution}
+
+ This appendix provides a description of the functions that constitute the
+ RSiena package. This is intended as a quick reference or catalogue for the
[TRUNCATED]
To get the complete diff run:
svnlook diff /svnroot/rsiena -r 73
More information about the Rsiena-commits
mailing list