Revision 678 trunk/docs/manuals/admbre/admbre.tex
admbre.tex (revision 678)  

49  49  
50  50 
Important points are emphasized with a star 
51  51 
\begin{itemize} 
52 
\item[$\bigstar$] like this. 

52 
\item[$\bigstar$] like this.


53  53 
\end{itemize} 
54  54  
55  55 
Please submit all comments and complaints by email to 
...  ...  
119  119 
models. With \scAR, you can include random effects in your model. Examples of 
120  120 
such models include: 
121  121 
\begin{itemize} 
122 
\item Generalized linear mixed models (logistic and Poisson regression). 

123 
\item Nonlinear mixed models (growth curve models, pharmacokinetics). 

124 
\item State space models (nonlinear Kalman filters). 

125 
\item Frailty models in survival analysis. 

126 
\item Nonparametric smoothing. 

127 
\item Semiparametric modelling. 

128 
\item Frailty models in survival analysis. 

129 
\item Bayesian hierarchical models. 

130 
\item General nonlinear randomeffects models (fisheries catchatage 

131 
models). 

122 
\item Generalized linear mixed models (logistic and Poisson regression). 

123 
\item Nonlinear mixed models (growth curve models, pharmacokinetics). 

124 
\item State space models (nonlinear Kalman filters). 

125 
\item Frailty models in survival analysis. 

126 
\item Nonparametric smoothing. 

127 
\item Semiparametric modelling. 

128 
\item Frailty models in survival analysis. 

129 
\item Bayesian hierarchical models. 

130 
\item General nonlinear randomeffects models (fisheries catchatage models). 

132  131 
\end{itemize} 
133  132 
You formulate the likelihood function in a template file, using a language that 
134  133 
resembles \cplus. The file is compiled into an executable program (on Linux or 
...  ...  
159  158 
\subsection{The strengths of \scAR} 
160  159  
161  160 
\begin{itemize} 
162 
\item \textit{Flexibility:} You can fit a large variety of models within a 

163 
single framework. 

164 
\item \textit{Convenience:} Computational details are transparent. Your only 

165 
responsibility is to formulate the loglikelihood. 

166 
\item \textit{Computational efficiency:} \scAR\ is up to 50 times faster


167 
than is \scWinBUGS.


168 
\item \textit{Robustness:} With exact derivatives, you can fit highly nonlinear


169 
models. 

170 
\item \textit{Convergence diagnostic:} The gradient of the likelihood


171 
function provides a clear convergence diagnostic.


161 
\item \textit{Flexibility:} You can fit a large variety of models within a


162 
single framework.


163 
\item \textit{Convenience:} Computational details are transparent. Your only


164 
responsibility is to formulate the loglikelihood.


165 
\item \textit{Computational efficiency:} \scAR\ is up to 50 times faster than


166 
\scWinBUGS.


167 
\item \textit{Robustness:} With exact derivatives, you can fit highly


168 
nonlinear models.


169 
\item \textit{Convergence diagnostic:} The gradient of the likelihood function


170 
provides a clear convergence diagnostic.


172  171 
\end{itemize} 
173  172  
174  173 
\subsection{Program interface} 
175  174  
176  175 
\begin{itemize} 
177 
\item\textit{Model formulation}: You fill in a \cplusbased template using your


178 
favorite text editor. 

176 
\item\textit{Model formulation}: You fill in a \cplusbased template using


177 
your favorite text editor.


179  178  
180 
\item \textit{Compilation}: You turn your model into an executable program using


181 
a \cplus\ compiler (which you need to install separately).


179 
\item \textit{Compilation}: You turn your model into an executable program


180 
using a \cplus\ compiler (which you need to install separately).


182  181  
183 
\item\textit{Platforms}: Windows and Linux


182 
\item\textit{Platforms}: Windows and Linux.


184  183 
\end{itemize} 
185  184  
186  185 
\subsection{How to obtain \scAR} 
...  ...  
199  198 
to data: 
200  199  
201  200 
\begin{itemize} 
202 
\item Vectormatrix arithmetic and vectorized operations for common mathematical 

203 
functions. 

204 
\item Reading and writing vector and matrix objects to a file. 

205 
\item Fitting the model is a stepwise manner (with ``phases''), where more and 

206 
more parameters become active in the minimization. 

207 
\item Calculating standard deviations of arbitrary functions of the model 

208 
parameters by the ``delta method.'' 

209 
\item \scMCMC\ sampling around the posterior mode. 

201 
\item Vectormatrix arithmetic and vectorized operations for common 

202 
mathematical functions. 

203 
\item Reading and writing vector and matrix objects to a file. 

204 
\item Fitting the model is a stepwise manner (with ``phases''), where more and 

205 
more parameters become active in the minimization. 

206 
\item Calculating standard deviations of arbitrary functions of the model 
Also available in: Unified diff