导航
习题
Q1:Generalized Geometric Brownian Motion
题面:
Let
S
(
t
)
S(t)
S(t) be a positive stochastic process that satisfies
d
S
(
t
)
=
α
(
t
)
S
(
t
)
d
t
+
σ
(
t
)
S
(
t
)
d
W
(
t
)
dS(t)=\alpha(t)S(t)dt+\sigma(t)S(t)dW(t)
dS(t)=α(t)S(t)dt+σ(t)S(t)dW(t)
where
α
(
t
)
\alpha(t)
α(t) and
σ
(
t
)
\sigma(t)
σ(t) are processes adapted to the Filtration
F
(
t
)
,
t
≥
0
\mathcal{F}(t), t\geq0
F(t),t≥0,compute
S
(
t
)
S(t)
S(t) through computing
d
ln
(
S
(
t
)
)
d\ln(S(t))
dln(S(t)).
解析:
根据伊藤公式
ln
S
t
=
ln
S
0
+
∫
0
t
σ
s
d
w
s
+
∫
0
t
(
α
s
−
1
2
σ
s
2
)
d
s
\ln S_t=\ln S_0+\int_0^t\sigma_sdw_s+\int_0^t(\alpha_s-\frac{1}{2}\sigma_s^2)ds
lnSt=lnS0+∫0tσsdws+∫0t(αs−21σs2)ds
方程两边同时取
exp
\exp
exp指数得到
S
t
=
S
0
exp
(
∫
0
t
σ
s
d
w
s
+
∫
0
t
(
α
s
−
1
2
σ
s
2
)
d
s
)
S_t=S_0\exp(\int_0^t\sigma_sdw_s+\int_0^t(\alpha_s-\frac{1}{2}\sigma_s^2)ds)
St=S0exp(∫0tσsdws+∫0t(αs−21σs2)ds)
Q2:计算Brownian Motion高阶矩
题面:
Let
W
(
t
)
W(t)
W(t) be a Brownian motion for
0
≤
t
≤
T
0\leq t\leq T
0≤t≤T,Please find the expection
E
W
6
(
T
)
\mathbb{E}W^6(T)
EW6(T).
解析:
伊藤积分
I
(
t
)
=
∫
0
t
Δ
(
s
)
d
w
(
s
)
I(t)=\int_0^t\Delta(s)dw(s)
I(t)=∫0tΔ(s)dw(s)
期望
E
[
I
(
t
)
]
=
0
\mathbb{E}[I(t)]=0
E[I(t)]=0
方差
E
[
I
2
(
t
)
]
=
E
[
∫
0
t
Δ
2
(
u
)
d
u
]
=
∫
0
t
E
[
Δ
2
(
u
)
]
d
u
\mathbb{E}[I^2(t)]=\mathbb{E}[\int_0^t\Delta^2(u)du]=\int_0^t\mathbb{E}[\Delta^2(u)]du
E[I2(t)]=E[∫0tΔ2(u)du]=∫0tE[Δ2(u)]du
设
f
(
t
,
x
)
=
x
6
,
f
t
=
0
,
f
x
=
6
x
5
,
f
x
x
=
30
x
4
f(t, x)=x^6, f_t=0, f_x=6x^5, f_{xx}=30x^4
f(t,x)=x6,ft=0,fx=6x5,fxx=30x4
由伊藤公式
d
f
(
t
,
W
(
t
)
)
=
d
W
6
(
t
)
=
6
W
5
(
t
)
d
W
(
t
)
+
1
2
∗
30
W
4
(
t
)
d
t
=
6
W
5
(
t
)
d
W
(
t
)
+
15
W
4
(
t
)
d
t
df(t, W(t))=dW^6(t)=6W^5(t)dW(t)+\frac{1}{2}*30W^4(t)dt=6W^5(t)dW(t)+15W^4(t)dt
df(t,W(t))=dW6(t)=6W5(t)dW(t)+21∗30W4(t)dt=6W5(t)dW(t)+15W4(t)dt
在
[
0
,
T
]
[0, T]
[0,T]上积分
W
6
(
T
)
−
W
6
(
0
)
=
∫
0
T
6
W
5
(
t
)
d
W
(
t
)
⏟
伊藤积分期望为0
+
∫
0
T
15
W
4
(
t
)
d
t
W^6(T)-W^6(0)=\underbrace{\int_0^T6W^5(t)dW(t)}_{\text{伊藤积分期望为0}}+\int_0^T15W^4(t)dt
W6(T)−W6(0)=伊藤积分期望为0
∫0T6W5(t)dW(t)+∫0T15W4(t)dt
两边同时求数学期望
E
[
W
6
(
T
)
]
=
15
∫
0
T
E
W
4
(
t
)
d
t
\mathbb{E}[W^6(T)]=15\int_0^T\mathbb{E}W^4(t)dt
E[W6(T)]=15∫0TEW4(t)dt
根据性质,如果
X
∼
N
(
0
,
σ
2
)
,
E
[
X
4
]
=
3
σ
4
X\sim \mathcal{N}(0, \sigma^2),\mathbb{E}[X^4]=3\sigma^4
X∼N(0,σ2),E[X4]=3σ4.
∵
W
(
t
)
∼
N
(
0
,
t
)
∴
E
[
W
4
(
t
)
]
=
3
t
2
\because W(t)\sim \mathcal{N}(0, t)\therefore \mathbb{E}[W^4(t)]=3t^2
∵W(t)∼N(0,t)∴E[W4(t)]=3t2
代入积分方程得到
E
[
W
6
(
T
)
]
=
15
T
3
\mathbb{E}[W^6(T)]=15T^3
E[W6(T)]=15T3
Q3:Vasicek Model
题面:
The Vasicek interest rate stochastic differential equation is
d
R
(
t
)
=
(
α
−
β
R
(
t
)
)
d
t
+
σ
d
W
(
t
)
dR(t)=(\alpha-\beta R(t))dt+\sigma dW(t)
dR(t)=(α−βR(t))dt+σdW(t)
where
α
,
β
\alpha, \beta
α,β and
σ
\sigma
σ are positive constants. Please compute
R
(
t
)
R(t)
R(t) through computing
d
(
e
β
t
R
(
t
)
)
d(e^{\beta t}R(t))
d(eβtR(t)).
解析:
令
f
(
t
,
x
)
=
e
β
t
x
,
f
t
=
x
β
e
β
t
,
f
x
=
e
β
t
,
f
x
x
=
0
f(t, x)=e^{\beta t}x, f_t=x\beta e^{\beta t}, f_x=e^{\beta t}, f_{xx}=0
f(t,x)=eβtx,ft=xβeβt,fx=eβt,fxx=0
计算微分并代入
R
(
t
)
R(t)
R(t)表达式得
d
(
e
β
t
R
(
t
)
)
=
d
(
f
(
t
,
R
(
t
)
)
)
=
R
(
t
)
β
e
β
t
d
t
+
e
β
t
d
R
(
t
)
=
α
e
β
t
d
t
+
σ
e
β
t
d
W
(
t
)
\begin{aligned} d(e^{\beta t}R(t))&=d(f(t, R(t)))=R(t)\beta e^{\beta t}dt+e^{\beta t}dR(t)\\ &=\alpha e^{\beta t}dt+\sigma e^{\beta t} dW(t) \end{aligned}
d(eβtR(t))=d(f(t,R(t)))=R(t)βeβtdt+eβtdR(t)=αeβtdt+σeβtdW(t)
在
[
0
,
t
]
[0, t]
[0,t]上对方程两边进行积分
e
β
t
(
R
(
t
)
)
−
R
(
0
)
=
∫
0
t
α
e
β
s
d
s
+
∫
0
t
σ
e
β
s
d
W
(
s
)
e^{\beta t}(R(t))-R(0)=\int_0^t\alpha e^{\beta s}ds+\int_0^t\sigma e^{\beta s} dW(s)
eβt(R(t))−R(0)=∫0tαeβsds+∫0tσeβsdW(s)
移项积分求解得到
R
(
t
)
=
e
−
β
t
R
(
0
)
+
α
β
(
1
−
e
−
β
t
)
+
∫
0
t
σ
e
−
β
(
t
−
s
)
d
W
(
s
)
R(t)=e^{-\beta t}R(0)+\frac{\alpha}{\beta}(1-e^{-\beta t})+\int_0^t\sigma e^{-\beta(t-s)}dW(s)
R(t)=e−βtR(0)+βα(1−e−βt)+∫0tσe−β(t−s)dW(s)
可以计算一些性质,利用伊藤积分期望为0
E
[
R
(
t
)
]
=
e
−
β
t
R
(
0
)
+
α
β
(
1
−
e
−
β
t
)
\mathbb{E}[R(t)]=e^{-\beta t}R(0)+\frac{\alpha}{\beta}(1-e^{-\beta t})
E[R(t)]=e−βtR(0)+βα(1−e−βt)
计算极限
lim
t
→
∞
E
[
R
(
t
)
]
=
α
β
\lim_{t\to\infty} \mathbb{E}[R(t)]=\frac{\alpha}{\beta}
t→∞limE[R(t)]=βα
可以看出
R
(
t
)
R(t)
R(t)具有均值回归的特性.
Q4:Stratonovich Integral
题面:
Let
W
(
t
)
,
0
≤
t
≤
T
W(t), 0\leq t\leq T
W(t),0≤t≤T be a Brownian motion. Let
T
T
T be a fixed positive number and let
Π
=
{
t
0
,
t
1
,
…
,
t
n
}
\Pi=\{t_0, t_1, \dots, t_n\}
Π={t0,t1,…,tn} be a partition of
[
0
,
T
]
[0, T]
[0,T]. For each
j
j
j, define
t
j
∗
=
t
j
+
t
j
+
1
2
t_j^*=\frac{t_j+t_{j+1}}{2}
tj∗=2tj+tj+1 to be the midpoint of the internal
[
t
j
,
t
j
+
1
]
[t_j, t_{j+1}]
[tj,tj+1].
(1) Define the half-sample quadratic variation corresponding to
Π
\Pi
Π to be
Q
Π
/
2
=
∑
j
=
0
n
−
1
[
W
(
t
j
∗
)
−
W
(
t
j
)
]
2
Q_{\Pi/2}=\sum_{j=0}^{n-1}[W(t_j^*)-W(t_j)]^2
QΠ/2=j=0∑n−1[W(tj∗)−W(tj)]2
Show that
Q
Π
/
2
Q_{\Pi/2}
QΠ/2 has limit
1
2
T
\frac{1}{2}T
21T as
∥
Π
∥
→
0
\lVert\Pi\rVert\to 0
∥Π∥→0
解析:
计算
E
[
Q
Π
/
2
]
\mathbb{E}[Q_{\Pi/2}]
E[QΠ/2]
E
[
Q
Π
/
2
]
=
∑
j
=
0
n
−
1
E
[
W
(
t
j
∗
)
−
W
(
t
j
)
]
2
\mathbb{E}[Q_{\Pi/2}]=\sum_{j=0}^{n-1}\mathbb{E}[W(t_j^*)-W(t_j)]^2
E[QΠ/2]=j=0∑n−1E[W(tj∗)−W(tj)]2
令
D
j
=
W
(
t
j
∗
)
−
W
(
t
j
)
D_j= W(t_j^*)-W(t_j)
Dj=W(tj∗)−W(tj)
D
j
=
W
(
t
j
∗
)
−
W
(
t
j
)
∼
N
(
0
,
t
j
∗
−
t
j
)
=
N
(
0
,
t
j
+
1
−
t
j
2
)
D_j=W(t_j^*)-W(t_j)\sim \mathcal{N}(0, t_j^*-t_j)=\mathcal{N}(0, \frac{t_{j+1}-t_j}{2})
Dj=W(tj∗)−W(tj)∼N(0,tj∗−tj)=N(0,2tj+1−tj)
∴
\therefore
∴
E
[
Q
Π
/
2
]
=
∑
j
=
0
n
−
1
E
[
D
j
2
]
\mathbb{E}[Q_{\Pi/2}]=\sum_{j=0}^{n-1}\mathbb{E}[D_j^2]
E[QΠ/2]=j=0∑n−1E[Dj2]
∴
\therefore
∴
V
[
D
j
]
=
E
[
D
j
2
]
−
E
[
D
j
]
2
=
t
j
+
1
−
t
j
2
\mathbb{V}[D_j]=\mathbb{E}[D_j^2]-\mathbb{E}[D_j]^2=\frac{t_{j+1}-t_j}{2}
V[Dj]=E[Dj2]−E[Dj]2=2tj+1−tj
∴
\therefore
∴
E
[
Q
Π
/
2
]
=
∑
j
=
0
n
−
1
t
j
+
1
−
t
j
2
=
t
n
−
t
0
2
=
T
2
\mathbb{E}[Q_{\Pi/2}]=\sum_{j=0}^{n-1}\frac{t_{j+1}-t_j}{2}=\frac{t_n-t_0}{2}=\frac{T}{2}
E[QΠ/2]=j=0∑n−12tj+1−tj=2tn−t0=2T
计算方差,根据布朗运动的增量独立性
V
[
Q
Π
/
2
]
=
V
[
∑
j
=
0
n
−
1
D
j
2
]
=
∑
j
=
0
n
−
1
V
[
D
j
2
]
\mathbb{V}[Q_{\Pi/2}]=\mathbb{V}[\sum_{j=0}^{n-1}D_j^2]=\sum_{j=0}^{n-1}\mathbb{V}[D_j^2]
V[QΠ/2]=V[j=0∑n−1Dj2]=j=0∑n−1V[Dj2]
∵
\because
∵
V
[
D
j
2
]
=
E
[
D
j
4
]
−
E
2
[
D
j
2
]
=
3
4
(
t
j
+
1
−
t
j
)
2
−
1
4
(
t
j
+
1
−
t
j
)
2
=
1
2
(
t
j
+
1
−
t
j
)
2
\mathbb{V}[D_j^2]=\mathbb{E}[D_j^4]-\mathbb{E}^2[D_j^2]=\frac{3}{4}(t_{j+1}-t_j)^2-\frac{1}{4}(t_{j+1}-t_j)^2=\frac{1}{2}(t_{j+1}-t_j)^2
V[Dj2]=E[Dj4]−E2[Dj2]=43(tj+1−tj)2−41(tj+1−tj)2=21(tj+1−tj)2
∴
\therefore
∴
V
[
Q
Π
/
2
]
=
1
2
∑
j
=
0
n
−
1
(
t
j
+
1
−
t
j
)
2
\mathbb{V}[Q_{\Pi/2}]=\frac{1}{2}\sum_{j=0}^{n-1}(t_{j+1}-t_j)^2
V[QΠ/2]=21j=0∑n−1(tj+1−tj)2
当
max
(
t
j
+
1
−
t
j
)
=
∥
Π
∥
→
0
\max(t_{j+1}-t_j)=\lVert\Pi\rVert\to 0
max(tj+1−tj)=∥Π∥→0时
V
[
Q
Π
/
2
]
=
1
2
∑
j
=
0
n
−
1
(
t
j
+
1
−
t
j
)
(
t
j
+
1
−
t
j
)
≤
1
2
∥
Π
∥
∑
j
=
0
n
−
1
(
t
j
+
1
−
t
j
)
=
1
2
∥
Π
∥
T
→
0
\mathbb{V}[Q_{\Pi/2}]=\frac{1}{2}\sum_{j=0}^{n-1}(t_{j+1}-t_j)(t_{j+1}-t_j)\leq \frac{1}{2}\lVert\Pi\rVert\sum_{j=0}^{n-1}(t_{j+1}-t_j)=\frac{1}{2}\lVert\Pi\rVert T\to 0
V[QΠ/2]=21j=0∑n−1(tj+1−tj)(tj+1−tj)≤21∥Π∥j=0∑n−1(tj+1−tj)=21∥Π∥T→0
(2) Define the Stratonovich Integral of
W
(
t
)
W(t)
W(t) with respect to
W
(
t
)
W(t)
W(t) to be
∫
0
T
W
(
t
)
∘
d
W
(
t
)
=
lim
∥
Π
∥
→
0
∑
j
=
0
n
−
1
W
(
t
j
∗
)
[
W
(
t
j
+
1
)
−
W
(
t
j
)
]
\int_0^TW(t)\circ dW(t)=\lim_{\lVert\Pi\rVert\to 0}\sum_{j=0}^{n-1}W(t_j^*)[W(t_{j+1})-W(t_j)]
∫0TW(t)∘dW(t)=∥Π∥→0limj=0∑n−1W(tj∗)[W(tj+1)−W(tj)]
证明
∫
0
T
W
(
t
)
∘
d
W
(
t
)
=
1
2
W
2
(
T
)
\int_0^TW(t)\circ dW(t)=\frac{1}{2}W^2(T)
∫0TW(t)∘dW(t)=21W2(T)
解析:
考虑从伊藤积分(区间补全)凑出Stratonovich Integral
Ito Intrgral
I
=
∫
0
T
W
(
t
)
d
W
(
t
)
=
∑
j
=
0
n
−
1
W
(
t
j
)
[
W
(
t
j
∗
)
−
W
(
t
j
)
]
+
∑
j
=
0
n
−
1
W
(
t
j
∗
)
[
W
(
t
j
+
1
)
−
W
(
t
j
∗
)
]
I=\int_0^TW(t)dW(t)=\sum_{j=0}^{n-1}W(t_j)[W(t_j^*)-W(t_j)]+\sum_{j=0}^{n-1}W(t_j^*)[W(t_{j+1})-W(t_j^*)]
I=∫0TW(t)dW(t)=j=0∑n−1W(tj)[W(tj∗)−W(tj)]+j=0∑n−1W(tj∗)[W(tj+1)−W(tj∗)]
Stratonovich Integral
S
=
I
−
2
∑
j
=
0
n
−
1
W
(
t
j
)
W
(
t
j
∗
)
+
∑
j
=
0
n
−
1
W
2
(
t
j
)
+
∑
j
=
0
n
−
1
W
2
(
t
j
∗
)
=
I
+
∑
j
=
0
n
−
1
[
W
(
t
j
)
−
W
(
t
j
∗
)
]
2
=
由
(
1
)
结
果
I
+
T
2
\begin{aligned} S&=I-2\sum_{j=0}^{n-1}W(t_j)W(t_j^*)+\sum_{j=0}^{n-1}W^2(t_j)+\sum_{j=0}^{n-1}W^2(t_j^*)\\ &=I+\sum_{j=0}^{n-1}[W(t_j)-W(t_j^*)]^2\\ &\xlongequal{由(1)结果}I+\frac{T}{2} \end{aligned}
S=I−2j=0∑n−1W(tj)W(tj∗)+j=0∑n−1W2(tj)+j=0∑n−1W2(tj∗)=I+j=0∑n−1[W(tj)−W(tj∗)]2由(1)结果I+2T
∴
\therefore
∴
∫
0
T
W
(
t
)
∘
d
W
(
t
)
=
∫
0
T
W
(
t
)
d
W
(
t
)
+
T
2
\int_0^TW(t)\circ dW(t)=\int_0^TW(t)dW(t)+\frac{T}{2}
∫0TW(t)∘dW(t)=∫0TW(t)dW(t)+2T
求伊藤积分
令
f
(
t
,
x
)
=
x
2
,
f
t
=
0
,
f
x
=
2
x
,
f
x
x
=
2
f(t, x)=x^2, f_t=0, f_x=2x, f_{xx}=2
f(t,x)=x2,ft=0,fx=2x,fxx=2
d
f
(
t
,
W
(
t
)
)
=
d
[
W
2
(
t
)
]
=
2
W
(
t
)
d
W
(
t
)
+
1
2
∗
2
∗
d
t
=
2
W
(
t
)
d
W
(
t
)
+
d
t
df(t, W(t))=d[W^2(t)]=2W(t)dW(t)+\frac{1}{2}*2*dt=2W(t)dW(t)+dt
df(t,W(t))=d[W2(t)]=2W(t)dW(t)+21∗2∗dt=2W(t)dW(t)+dt
∴
\therefore
∴
∫
0
T
W
2
(
t
)
d
W
(
t
)
=
∫
0
T
2
W
(
t
)
d
W
(
t
)
+
∫
0
T
d
t
W
2
(
T
)
−
W
2
(
0
)
=
2
∫
0
T
W
(
t
)
d
W
(
t
)
+
∫
0
T
d
t
∫
0
T
W
(
t
)
d
W
(
t
)
=
W
2
(
T
)
−
T
2
\begin{aligned} &\int_0^TW^2(t)dW(t)=\int_0^T2W(t)dW(t)+\int_0^Tdt \\ &W^2(T)-W^2(0)=2\int_0^TW(t)dW(t)+\int_0^Tdt\\ &\int_0^TW(t)dW(t)=\frac{W^2(T)-T}{2} \end{aligned}
∫0TW2(t)dW(t)=∫0T2W(t)dW(t)+∫0TdtW2(T)−W2(0)=2∫0TW(t)dW(t)+∫0Tdt∫0TW(t)dW(t)=2W2(T)−T
代入计算可以得到
∫
0
T
W
(
t
)
∘
d
W
(
t
)
=
∫
0
T
W
(
t
)
d
W
(
t
)
+
T
2
=
W
2
(
T
)
2
\int_0^TW(t)\circ dW(t)=\int_0^TW(t)dW(t)+\frac{T}{2}=\frac{W^2(T)}{2}
∫0TW(t)∘dW(t)=∫0TW(t)dW(t)+2T=2W2(T)