257 lines
13 KiB
TeX
257 lines
13 KiB
TeX
\documentclass[landscape,final,a0paper,fontscale=0.285]{baposter}
|
|
|
|
\usepackage{calc}
|
|
\usepackage{graphicx}
|
|
\usepackage{amsmath}
|
|
\usepackage{amssymb}
|
|
\usepackage{relsize}
|
|
\usepackage{multirow}
|
|
\usepackage{rotating}
|
|
\usepackage{bm}
|
|
\usepackage{url}
|
|
\usepackage[francais]{babel}
|
|
\usepackage{graphicx}
|
|
\usepackage{multicol}
|
|
\usepackage{amsmath}
|
|
\usepackage{amsfonts}
|
|
\usepackage[utf8]{inputenc}
|
|
\usepackage[T1]{fontenc}
|
|
%\usepackage{times}
|
|
%\usepackage{helvet}
|
|
%\usepackage{bookman}
|
|
\usepackage{palatino}
|
|
|
|
\newcommand{\captionfont}{\footnotesize}
|
|
|
|
\graphicspath{{images/}{../images/}}
|
|
\usetikzlibrary{calc}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%%% Some math symbols used in the text
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Multicol Settings
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\setlength{\columnsep}{1.5em}
|
|
\setlength{\columnseprule}{0mm}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Save space in lists. Use this after the opening of the list
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\newcommand{\compresslist}{%
|
|
\setlength{\itemsep}{1pt}%
|
|
\setlength{\parskip}{0pt}%
|
|
\setlength{\parsep}{0pt}%
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Begin of Document
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
|
|
\begin{document}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Here starts the poster
|
|
%%%---------------------------------------------------------------------------
|
|
%%% Format it to your taste with the options
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Define some colors
|
|
|
|
%\definecolor{lightblue}{cmyk}{0.83,0.24,0,0.12}
|
|
\definecolor{lightblue}{rgb}{0.145,0.6666,1}
|
|
\hyphenation{resolution occlusions}
|
|
%%
|
|
\begin{poster}%
|
|
% Poster Options
|
|
{
|
|
% Show grid to help with alignment
|
|
grid=false,
|
|
% Column spacing
|
|
colspacing=1em,
|
|
% Color style
|
|
bgColorOne=white,
|
|
bgColorTwo=white,
|
|
borderColor=lightblue,
|
|
headerColorOne=black,
|
|
headerColorTwo=lightblue,
|
|
headerFontColor=white,
|
|
boxColorOne=white,
|
|
boxColorTwo=lightblue,
|
|
% Format of textbox
|
|
textborder=roundedleft,
|
|
% Format of text header
|
|
eyecatcher=true,
|
|
headerborder=closed,
|
|
headerheight=0.1\textheight,
|
|
% textfont=\sc, An example of changing the text font
|
|
headershape=roundedright,
|
|
headershade=shadelr,
|
|
headerfont=\Large\bf\textsc, %Sans Serif
|
|
textfont={\setlength{\parindent}{1.5em}},
|
|
boxshade=plain,
|
|
% background=shade-tb,
|
|
background=plain,
|
|
linewidth=2pt
|
|
}
|
|
% Eye Catcher
|
|
{\includegraphics[height=5em]{images/ullong.pdf}}
|
|
% Title
|
|
{\bf\textsc{Distribution de Laplace asymétrique généralisée}\vspace{0.5em}}
|
|
% Authors
|
|
{\textsc{François Pelletier, Andrew Luong} \\ \small{École d'actuariat, Université Laval}}
|
|
% University logo
|
|
{% The makebox allows the title to flow into the logo, this is a hack because of the L shaped logo.
|
|
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Now define the boxes that make up the poster
|
|
%%%---------------------------------------------------------------------------
|
|
%%% Each box has a name and can be placed absolutely or relatively.
|
|
%%% The only inconvenience is that you can only specify a relative position
|
|
%%% towards an already declared box. So if you have a box attached to the
|
|
%%% bottom, one to the top and a third one which should be in between, you
|
|
%%% have to specify the top and bottom boxes before you specify the middle
|
|
%%% box.
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%
|
|
% A coloured circle useful as a bullet with an adjustably strong filling
|
|
\newcommand{\colouredcircle}{%
|
|
\tikz{\useasboundingbox (-0.2em,-0.32em) rectangle(0.2em,0.32em); \draw[draw=black,fill=lightblue,line width=0.03em] (0,0) circle(0.18em);}}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Introduction}{name=problem,column=0,row=0}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{On définit le prix au temps $t$ d'un titre financier $S(t)$ et le rendement cumulé sur ce titre
|
|
\begin{equation*}L_t = R_1 + \ldots + R_{t} = \log S(t) - \log S(0). \end{equation*}
|
|
On veut modéliser $R_t$ par la distribution de Laplace asymétrique généralisée, aussi appelée variance-gamma. L'estimation paramétrique se fera à partir de la méthode GMM.
|
|
On pourra ainsi obtenir la distribution de $L_t$ par convolution, afin d'évaluer le prix d'options européennes.}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Simulation}{name=contribution,column=1,row=0,bottomaligned=problem}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{Le processus de Laplace $Y(t)$ peut être représenté comme la différence de deux processus gamma $G(t)$ \cite{kotz2001laplace}.
|
|
\begin{equation*} Y \stackrel{d}{=} \theta + \frac{\sigma}{\sqrt{2}} \left( \frac{1}{\kappa} G_1 - \kappa G_2 \right)
|
|
\end{equation*} où $ G_1,G_2 \sim \Gamma\left(\alpha=\tau,\beta=1 \right)$.
|
|
Il suffit donc de simuler deux réalisations de cette variable
|
|
aléatoire pour obtenir une réalisation de la distribution Laplace
|
|
asymétrique généralisée.
|
|
}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Estimation du modèle}{name=results,column=2,span=1,row=0}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{On utilise les conditions de moment basées sur l'espérance et la variance
|
|
\begin{equation*}
|
|
g(Y_t,\theta) =
|
|
\left[\begin{array}[]{c}
|
|
\bar{Y} - \left(\theta+\mu\tau\right) \\
|
|
\frac{1}{T}\sum_{t=1}^T \left(Y_t - \bar{Y} \right)^2 - \tau\left(\sigma^2+\mu^2\right)
|
|
\end{array}\right]
|
|
\end{equation*}
|
|
La matrice optimale $W$ est estimée en utilisant la méthode du GMM itératif \cite{hall2005generalized}:
|
|
\begin{align*}
|
|
\hat{W}_{(1)} &= \bigg(\frac{1}{T}\sum_{t=1}^T g(Y_t,\hat\theta_{(0)})g(Y_t,\hat\theta_{(0)})^{\prime}\bigg)^{-1} \\
|
|
\hat{W}_{(i)} &= \bigg(\frac{1}{T}\sum_{t=1}^T g(Y_t,\hat\theta_{(i-1)})g(Y_t,\hat\theta_{(i-1)})^{\prime}\bigg)^{-1}
|
|
\end{align*}}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Graphique}{name=references,column=3,span=1,bottomaligned=results}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\includegraphics[width=64mm,height=56mm]{GraphiqueAffiche1}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Processus de Laplace}{name=references,column=0,above=bottom}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{Le processus de Laplace $Y(t)$ est un processus de Lévy subordonné, c'est-à-dire qu'il est formé de deux processus de Lévy, l'un qui détermine l'amplitude des évènements (sauts) et l'autre qui détermine le temps entre les évènements.
|
|
Le premier est un processus de Wiener $W(t+\tau)-W(t) \sim N(0,\tau\sigma^2)$ et le second un processus gamma $G(t+\tau)-G(t) \sim \Gamma(\tau,\nu\tau)$.
|
|
\begin{equation*}Y(t) = W(G(t))\end{equation*}}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Méthode GMM}{name=questions,column=1,span=1,aligned=references,above=bottom}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
La méthode GMM a pour objectif d'estimer les paramètres $\theta$ d'une distribution en minimisant une norme quadratique
|
|
\begin{equation*}
|
|
\|\hat{m}\left(\theta\right) \|_W^2 = \hat{m}\left(\theta\right)^{\prime} W \hat{m}\left(\theta\right)
|
|
\end{equation*}
|
|
pondérée par une matrice définie positive $W$ et où $\hat{m}\left(\theta\right) = \left(\frac{1}{T}\sum_{t=1}^T g(Y_t,\theta)\right)$ est la moyenne empirique des conditions de moment $g(Y_t,\theta)$.
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Références}{name=references,column=2,span=2,aligned=references,above=bottom}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\bibliographystyle{plain} % (uses file "plain.bst")
|
|
\bibliography{biblio} % expects file "myrefs.bib"
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Prix d'options}{name=speed,column=2,span=2,row=0,below=results,above=references}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{L'évaluation du prix d'un option de vente européenne équivaut à calculer la valeur espérée de la réclamation contingente définie par le contrat
|
|
\begin{equation}
|
|
P(Y(t),T-t) = B(t,T) \int_{0}^{K} (K-Y(T)) d\hat{F}_t(Y(T))
|
|
\end{equation}
|
|
où $B(t,T)$ est la valeur d'une obligation zéro-coupon au taux sans
|
|
risque $r$ d'échéance $T$ et $K$ le prix d'exercice. Cette évaluation doit de faire sous une mesure neutre au risque, selon laquelle les investisseurs n'exigent pas une prime de risque. Cette mesure doit répondre à la propriété martingale. Pour la distribution étudiée, ceci se fait par une modification du paramètre de dérive $\theta$ en $\theta^{\star}$
|
|
\begin{align*}
|
|
e^{rt} = E\left[exp(L_t) \right] = M_{Y_1+\ldots+Y_t}(1) = (\phi_{Y}(-i))^t = \left(\frac{e^{\theta^{\star}}}{(1-\mu-\sigma^2/2)^{\tau}} \right)^t \Rightarrow \theta^{\star} = r + \tau\log(1-\mu-\frac{\sigma^2}{2})
|
|
\end{align*}
|
|
On pourra ensuite évaluer la valeur du prix de l'option de vente avec la formule de Heston (1993) décrite dans \cite{epps2007pricing}
|
|
\begin{align*}
|
|
P(Y(t),T-t) &= B(t,T)K \int_{0}^{K} d{F}_t(Y(T)) - B(t,T)\int_{0}^{K} Y(T)\cdot d{F}_t(Y(T)) = B(t,T)K \hat{F}_t(K) - Y(t)\hat{G}_t(K)
|
|
\end{align*}
|
|
où $\hat{G}_t(K)$ est la fonction de répartition de la transformée d'Esscher (h=1) de la mesure neutre au risque.
|
|
}}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Choix du modèle}{name=method,column=0,below=problem,bottomaligned=speed}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{Inspirés par les modèles Mandelbrot, Press et Praetz , Madan et Seneta \cite{madan1990variance} présentent un ensemble de caractéristiques essentielles pour un modèle de rendements financiers:}
|
|
\small{\begin{enumerate}
|
|
\item Distribution de $R_t$ ayant une queue longue
|
|
\item Distribution de $R_t$ ayant des moments finis
|
|
\item Processus en temps continu ayant des accroissements stationnaires et indépendants. Distribution des accroissement de même famille peu importe la longueur.
|
|
\item Extension multivariée afin de conserver la validité du modèle CAPM.
|
|
\end{enumerate}}
|
|
\small{Ce tableau décrit le respect des conditions pour les différents modèles étudiés par les auteurs}
|
|
\begin{center}
|
|
\begin{tabular}{|l|cccc|}
|
|
\hline
|
|
& \multicolumn{4}{c|}{\scriptsize{Conditions}} \\
|
|
\hline
|
|
\scriptsize{Modèle} & \scriptsize{1} & \scriptsize{2} & \scriptsize{3} & \scriptsize{4} \\
|
|
\hline
|
|
\scriptsize{Mandelbrot} & \scriptsize{X} & & & \scriptsize{X} \\
|
|
\scriptsize{Press} & \scriptsize{X} & \scriptsize{X} & \scriptsize{X} & \scriptsize{X} \\
|
|
\scriptsize{Praetz} & \scriptsize{X} & \scriptsize{X} & & \scriptsize{X} \\
|
|
\scriptsize{Madan et Seneta} & \scriptsize{X} & \scriptsize{X} & \scriptsize{X} & \scriptsize{X} \\
|
|
\hline
|
|
\end{tabular}
|
|
\end{center}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Fonction caractéristique}{name=background model,column=1,below=problem,bottomaligned=speed}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\small{On définit la variable aléatoire $Y$
|
|
\begin{equation*}
|
|
\label{eq:defvarY-GAL}
|
|
Y = \theta + \mu W + \sigma \sqrt{W} Z
|
|
\end{equation*}
|
|
La fonction caractéristique de la distribution de $Y$ peut être
|
|
obtenue en utilisant la formule de l'espérance conditionnelle.
|
|
\begin{align*}
|
|
\phi_Y&(t;\theta,\sigma,\mu,\tau) = E\left[e^{ity}\right] \\
|
|
&= E\left[E\left[e^{ity} | W \right] \right] \\
|
|
&= \int_0^{\infty} E \left[ e^{it(\theta + \mu w+\sigma\sqrt{w}Z)} \right] g(w) dw \\
|
|
&= e^{i\theta t}\int_0^{\infty} e^{ i\mu w t-\frac{\sigma^2t^2w}{2}} \times \frac{1}{\Gamma (\tau)} w^{\tau-1}e^{-w} \\
|
|
&= \frac{e^{i\theta t}}{\left(1+\frac{1}{2} \sigma^2 t^2 - i\mu t \right)^{\tau}}
|
|
\end{align*}
|
|
On retrouve la fonction de répartition par le théorème de Gil-Pelaez
|
|
\begin{equation*}
|
|
\label{eq:inversionfncaract}
|
|
F_Y(y) = \frac{1}{2} - \frac{1}{\pi}\int_{0}^{\infty} \frac{Im\left[e^{-ity}\phi_Y(t)\right]}{t} dt
|
|
\end{equation*}}
|
|
}
|
|
\end{poster}
|
|
|
|
\end{document}
|