-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathvorlesung09.tex
99 lines (73 loc) · 4.26 KB
/
vorlesung09.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
\part{Gesetze der großen Zahlen}
\section{Einführung}
Es handelt sich um Grenzwertsätze der Stochastik. Wir müssen uns hierfür erst einmal Konvergenzbegriffe überlegen.
Seien $X,X_n,\ n\in \mathbb{N}$ neue, $d$-dim. Zufallsvariablen auf einem W-Raum $(\Omega,\mathfrak{A},P)$.
Die Folge der $X_n$ konvergiert ``$p$-fast-sicher'' gegen $X$, wenn gilt
\begin{equation}
P\left( \lim_{n\rightarrow\infty}{X_n=X} \right) = P \left( \bigcap_{l\in\mathbb{N}>0}\bigcup_{m\in\mathbb{N}}\bigcap_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\} \right) = 1
\end{equation}
Hierbei sei $|\cdot |$ die euklidische Norm)
Schreibweise: \[ X_n \rightarrow X,\ P\text{-f.s.} \]
Die Folge der $X_n$ konvergiert stochastisch gegen $X$, wenn gilt
\begin{equation}
\forall \epsilon>0 : \lim_{n\rightarrow\infty}P\left(|X_n-X| \geq \epsilon\right) = 0
\end{equation}
Schreibweise \[ X_n \stackrel{P}{\rightarrow} X\]
\section{Bemerkungen}
\subsection{a}
\begin{equation}
\begin{split}
& X_n \rightarrow X P\text{-f.s.} &= \iff \text{Es gibt eine P-Nullmenge $N\in \mathfrak{A}$, (d.h. $P(N)=0$) mit der Eig., dass}\\
& & \lim_{n\rightarrow\infty}X_n(\omega)=X(\omega) \forall \omega \in N^c\\
"`\Rightarrow"' & \text{Wähle } N &=\left( \bigcap_{l\in\mathbb{N}>0}\bigcup_{m\in\mathbb{N}}\bigcap_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}\right)^c \\
& & = \bigcup_{l\in\mathbb{N}>0}\bigcap_{m\in\mathbb{N}}\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}\\
"`\Leftarrow"' & N^c & \subset \bigcap_{l\in\mathbb{N}>0}\bigcup_{m\in\mathbb{N}}\bigcap_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}
\end{split}
\end{equation}
\subsection{b}
$X_n \rightarrow X$ $P$-f.s. $\Rightarrow X_n \stackrel{P}{\rightarrow} X$.\\
Für $l\in\mathbb{N}$ ist
\begin{equation}
\begin{split}
& \bigcap_{m\in\mathbb{N}}\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\} &\subset \bigcup_{k\in\mathbb{N}>0}\bigcap_{m\in\mathbb{N}}\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{k}\}\\
\Rightarrow & = P\left(\bigcap_{m\in\mathbb{N}}\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}\right) &= 0
\end{split}
\end{equation}
% neuer Abschnitt?
Also
\begin{equation}
\begin{split}
0 &= P\left(\bigcap_{m\in\mathbb{N}}\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}\right) \\
&= \lim_{m\rightarrow\infty} P\left(P\left(\bigcup_{n=m}^{\infty}\{|X_n-X| \leq \frac{1}{n}\}\right)\right) \\
&\geq \limsup_{m\rightarrow\infty} P\left(|X_n-X| \leq \frac{1}{n}\right)
\end{split}
\end{equation}
\section{Chebyshevsches schwaches Gesetz der großen Zahlen}
Seien $X_n$ unabhängige, reelle zufallsvariablen, je mit derselben Verteilung. Es sei $E(|X_1|^2)<\infty$.
Sei $X_n=\frac{1}{n}\cdot \sum_{j=1}^{n}{X_j},\ n\in\mathbb{N}, \mu = E(X_1)$. Dann gilt % X quer?
\begin{equation}
X_n \stackrel{P}{\rightarrow}\mu % x quer
\end{equation}
\subsection{Beweis}
Mit der Chebyshevschen Ungleichung und $\mu=E(X_n)$. % x quer
\begin{equation}
P(|X_n-\mu|) \leq \frac{1}{\epsilon^2}Var(X_n)=\frac{1}{\epsilon^2\cdot n^2} \sum_{j=1}^n{Var(X_j)} = \frac{1}{\epsilon\cdot n}Var(X_1) \stackrel{\rightarrow}{(n,\ldots,\infty)} 0
\end{equation}
\section{Starkes Gesetz der großen Zahlen}
von Kolmogorov, 1933.
Seien $X_n$ unabhängige reelle Zufallsvariablen, ke mit derselben Verteilung mit endlichen Erwartungswert $\mu$. Dann gilt
\[ X_n\rightarrow\mu, P\text{-f.s.} \] % x quer
\subsection{Anwendung}
Berechnung von Integralen mittels Monte-Carlo-Simulation
\[g:R^k\rightarrow \mathbb{R}, Q \leq g \leq 1\]
Es sei \[ \int_{-\infty}^{+\infty}\cdots\int_{-\infty}^{+\infty}{g(x_1,\ldots,x_n)\cdot dx_1\cdots dx_k} = \mu < \infty \]
Es ser $f > 0 $ die Dichte eines $k$-dim. Zufallsvektors $X$.
Dann ist
\[ \mu = \int_{-\infty}^{+\infty}\cdots\int_{-\infty}^{+\infty}{\underbrace{\frac{g(x_1,\ldots,x_k}{f(x_1,\ldots,x_k)}}_{= h(x_1,\ldots,x_k)}\cdot f(x_1,\ldots,x_k)\cdot dx_1\cdots dx_k} = E(h(X)) \]
% neuer abschnitt
Seien $X_1,X_2,\ldots$ unabh., je mit derselben Vtl. wie $X$.
$Y_n=h(X_n), n \in \mathbb{N}$ die $X_n$ sind unabh. je mit derselben Vtl. mit $\mu = E(Y_1)$.
\[ \frac{1}{1}\cdot Y_n = \frac{1}{n}\cdot\sum_{j=1}^n{h(X_j}\rightarrow\mu, P-f.s. \] % y quer
Verschaffe Beobachtung $x1,\ldots,x_n$ von $X_1,\ldots,X_n$ und fasse
$\frac{1}{n}\cdot\sum_{j=1}^{n}h(x_j) $ als Approximation für $\mu =\int\int{g(x_1,\ldots,x_k)dx_1\cdots dx_k}$ % int von -inf +ingf
auf