Clase 8 - Densidades marginales, independencia y cambio de variables

fXY(X,Y)={2x⋅e−x20≤x,0≤Y≤X20ccf_{XY}(X,Y)=\begin{cases} 2x\cdot e^{-x^{2}} & 0\leq x,0\leq Y\leq X^{2} \\ 0 & cc \end{cases}

Calculamos la marginal:

FX(X)=∫−∞+∞fXY(X,Y) dY=∫−∞+∞2Xe−X2⋅1[0,+∞)(X)⋅1[0,X2](Y) dYF_{X}(X) = \int_{-\infty}^{+\infty} f_{XY}(X,Y) \, dY = \int_{-\infty}^{+\infty} 2Xe^{-X^{2}} \cdot \mathbb{1}_{[0,+\infty)} (X) \cdot \mathbb{1}_{[0,X^{2}]}(Y) \, dY 1[0,+∞](X)⋅∫0X2x⋅e−X2 dY=1[0,+∞)(X)⋅2X⋅e−X2X2\mathbb{1}_{[0,+\infty]}(X) \cdot \int_{0}^{X} 2x \cdot e^{-X^{2}} \, dY = \mathbb{1}_{[0,+\infty)}(X)\cdot2X\cdot e^{-X^{2}} X^{2} =fX(X)=2X3⋅e−X2⋅1[0,+∞)(X)=f_{X}(X) = 2 X^{3} \cdot e^{-X^{2}} \cdot \mathbb{1}_{[0,+\infty)} (X) fY(Y)=∫−∞+∞2X⋅e−X2⋅1[0,+∞)(X)⋅1[0,X2](Y) dxf_{Y}(Y)= \int_{-\infty}^{+\infty} 2X \cdot e^{-X^{2}} \cdot \mathbb{1}_{[0,+\infty)} (X) \cdot \mathbb{1}_{[0,X^{2}]} (Y) \, dx =∫0+∞2X⋅e−X2⋅1[Y,+∞)(X)⋅1[0,+∞)(X)⋅⋅1[0,+∞)(Y) dX=\int_{0}^{+\infty} 2X \cdot e^{-X^{2}} \cdot \mathbb{1}_{[\sqrt{ Y },+\infty)} (X) \cdot\mathbb{1}_{[0,+\infty)}(X)\cdot\cdot\mathbb{1}_{[0,+\infty)}(Y)\, dX =1[0,+∞)(Y)⋅∫max{Y,0}+∞2X⋅e−X2 dx=⏟u=x2du=2x1[0,+∞)(Y)⋅∫Y+∞e−u du=−e−u∣y+∞=e−y=\mathbb{1}_{[0,+\infty)}(Y) \cdot \int_{max\{ \sqrt{ Y },0 \}}^{+\infty} 2X \cdot e^{-X^{2}} \, dx \underbrace{ = }_{ \begin{array}{c} u = x^{2} \\ du=2x \end{array} } \mathbb{1}_{[0,+\infty)}(Y) \cdot\int_{Y}^{+\infty} e ^{-u} \, du = - e^{-u} \bigm| _{y}^{+\infty} = e ^{-y}

Por lo tanto

fX(y)=e−Y⋅1[0,+∞)(y)f_{X}(y) = e^{-Y} \cdot \mathbb{1}_{[0,+\infty)}(y)

Se usó que

1[0,X2)(Y)=1[Y,+∞)(X)⋅1[0,+∞)(Y)\mathbb{1}_{[0,X^{2})}(Y) = \mathbb{1}_{[\sqrt{ Y },+\infty)}(X) \cdot \mathbb{1}_{[0,+\infty)}(Y)

XX e YY son independientes? No, ya que f(X,Y)≠fX⋅fYf(X,Y) \neq f_{X} \cdot f_{Y}

 ∃ S⊆{(x,y):y>x2,y>0,x>0}\:\exists\:S\subseteq \{ (x,y): y> x^{2}, y> 0, x > 0 \}

tal que fxfy=e−y⋅1[0,+∞)(Y)⋅2x2⋅e−x2⋅1[0,+∞)(x)≠0f_{x} f_{y}= e^{-y} \cdot \mathbb{1}_{[0,+\infty)}(Y) \cdot 2 x^{2} \cdot e^{-x^{2}} \cdot\mathbb{1}_{[0,+\infty)}(x) \neq 0

  ⟹  fxfy∣S=0\implies f_{x}f_{y}\bigm| _{S} = 0
Z∼U[0,1]Z \sim \mathcal{U}[0,1] \quad

ZZ independiente de YY

P(4⋅Z≤Y)?,(Y,Z)\mathbb{P}(4\cdot Z \leq Y)?, \quad (Y,Z) fY,Z(Y,Z)=1[0,1](Z)⋅e−y⋅1[0,+∞)(y)f_{Y,Z} (Y,Z) = \mathbb{1}_{[0,1]}(Z) \cdot e^{-y} \cdot \mathbb{1}_{[0,+\infty)} (y) ∫Rf[Y,Z](y,z)R={(y,z):4z≤y}\int_{R} f_{[Y,Z]}(y,z) \quad \quad R=\{ (y,z) :4z \leq y \} ∫−∞+∞∫−∞y41[0,1)(z)⋅e−y⋅1[0,+∞](y) dz dy\int_{-\infty}^{+\infty} \int_{-\infty}^{ \frac{y}{4}} \mathbb{1}_{[0,1)}(z)\cdot e^{-y} \cdot \mathbb{1}_{[0,+\infty]} (y) \, dz \, dy ∫0+∞e−y∫0min{y4,1}1 dz dy+∫04e−y∫0y41 dz dy\int_{0}^{+\infty} e^{-y} \int_{0}^{min\left\{ \frac{y}{4},1 \right\}} 1 \, dz \, dy +\int_{0}^{4} e ^{-y} \int_{0}^{ \frac{y}{4}} 1 \, dz \, dy ∫4+∞e−y∫011 dz dy+∫04e−y∫0y41 dz dy\int_{4}^{+\infty} e^{-y} \int_{0}^{1} 1 \, dz \, dy +\int_{0}^{4} e ^{-y} \int_{0}^{ \frac{y}{4}} 1 \, dz \, dy −e−y∣4+∞+∫04e−yy4 dy- e^{-y} \bigm| _{4}^{+\infty} + \int_{0}^{4} e^{-y} \frac{y}{4} \, dy e−4+… partes.e^{-4} + \dots \text{ partes.}
A∼E(15)B∼A24A independiente de BA \sim \mathcal{E}\left( \frac{1}{5} \right) \quad \quad B \sim \frac{A^{2}}{4} \quad \quad A \text{ independiente de } B h(t)={00<t≤222<t≤656<th(t) = \begin{cases} 0 & 0< t\leq 2 \\ 2 & 2<t \leq 6 \\ 5 & 6<t \end{cases}

¿fB?f_{B}?

Recordemos: XX variable aleatoria. Y∼g(X)Y \sim g(X)

fY(y)=fX(g−1(y))⋅∣g−1(y)′∣f_{Y}(y)= f_{X}(g ^{-1}(y)) \cdot |g ^{-1}(y)'| P(X∈(a,b))=1\mathbb{P}(X \in (a,b)) = 1

gg sea estrictamente creciente o decreciente en (a,b)(a,b)

X∼E(15)g(X)=X24B∼g(X)X \sim\mathcal{E}\left( \frac{1}{5} \right) \quad \quad g(X) = \frac{X^{2}}{4} \quad \quad B \sim g(X)

g−1(x)=4⋅xg ^{-1}(x) = \sqrt{ 4\cdot x }

fB(y)=5⋅e−4y⋅1[0,+∞)(4y)⋅∣12⋅14y∣f_{B}(y) = 5\cdot e ^{-\sqrt{ 4y }} \cdot \mathbb{1}_{[0,+\infty)}(\sqrt{ 4y}) \cdot | \frac{1}{2}\cdot \frac{1}{\sqrt{ 4y }}|

X∼h(A)X \sim h(A) Y∼h(B)Y\sim h(B) Calcular P(X,Y)(X,Y)P_{(X,Y)} (X,Y)

Notar que XX e YY son independientes.

fX(x)=1[0,2](x)⋅P(A∈[0,2])+1(2,6]⋅P(A∈[2,6])++1(6,+∞)(x)⋅P(A∈[6,+∞])\begin{array}{c} \displaystyle f_{X}(x) = \mathbb{1}_{[0,2]}(x) \cdot \mathbb{P}(A \in [0,2])+\mathbb{1}_{(2,6]} \cdot \mathbb{P}(A \in[2,6])+ \\ +\mathbb{1}_{(6,+\infty)} (x)\cdot \mathbb{P}(A \in [6,+\infty]) \end{array}

Esto último chequear, el profe lo borró

Llamemos

P0=P(A∈[0,2])P2=P(A∈[2,6])P5=P(A∈[6,+∞])\begin{array}{c} P_{0}=\mathbb{P}(A \in [0,2]) \\ P_{2}=\mathbb{P}(A \in [2,6]) \\ P_{5}=\mathbb{P}(A \in [6,+\infty]) \end{array} Fx(t)={0t<0P00≤t<2P2+P52≤t<515≤tF_{x}(t) = \begin{cases} 0 & t<0 \\ P_{0} & 0\leq t<2 \\ P_{2}+P_{5} & 2\leq t<5 \\ 1 & 5\leq t \end{cases} px(0)=P0py(0)=q0px(2)=P2py(2)=q2px(5)=P5py(5)=q5\begin{matrix} p_{x}(0)=P_{0} & p_{y}(0)=q_{0} \\ p_{x}(2)=P_{2} & p_{y}(2)=q_{2} \\ p_{x}(5)=P_{5} & p_{y}(5)=q_{5} \end{matrix} P(X,Y)(i,j)=pi⋅pji,j∈{0,2,5}P_{(X,Y)}(i,j)=p_{i}\cdot p_{j}\quad i,j \in \{ 0,2,5 \}

Falta integrar y calcular las p′sp's y q′sq's


X⃗=(X,Y)\vec{X}= (X,Y) \quad fX⃗(x,y)={10<x,0<y<ex0ccf_{\vec{X}}(x,y) = \begin{cases} 1 & 0<x,0<y<e^{x} \\ 0 & cc \end{cases}

a)

fX(x)=e−x⋅1[0,+∞)(x)f_{X}(x)=e^{-x} \cdot \mathbb{1}_{[0,+\infty)}(x) X∼E(1)X \sim\mathcal{E}(1) fY(y)=∫−∞+∞1[0,+∞)(x)⋅1[0,e−x](y) dxf_{Y}(y) = \int_{-\infty}^{+\infty} \mathbb{1}_{[0,+\infty)}(x) \cdot \mathbb{1}_{[0,e^{-x} ]}(y) \, dx 1[0,e−x](y)=1[−∞,−ln⁡(y)](x)⋅1[0,1](y)\mathbb{1}_{[0,e^{-x} ]}(y) = \mathbb{1}_{[-\infty ,- \ln(y) ]}(x) \cdot \mathbb{1}_{[0,1]}(y)   ⟹  fY(y)=1[0,1](y)⋅∫0−ln⁡(y)1 dx=−ln⁡(y)⋅1[0,1](y)\implies f_{Y}(y)=\mathbb{1}_{[0,1 ]}(y) \cdot \int_{0}^{- \ln(y)} 1 \, dx = -\ln(y) \cdot \mathbb{1}_{[0,1]} (y)

b)

Ye−X∼A[0,1]Y es independiente de X\frac{Y}{e^{-X} }\sim\mathcal{A}[0,1] \quad \quad Y \text{ es independiente de } X Y⃗=(U,V)U=XV=Ye−X\vec{Y}=(U,V)\quad \quad U=X\quad \quad V= \frac{Y}{e^{-X} } g(X,Y)=(X,Ye−X)g(X,Y) = \left( X, \frac{Y}{e^{-X} } \right) g⃗(u,v)=(u,v⋅e−u)\vec{g}(u,v)= (u,v\cdot e^{-u} ) fy⃗(u,v)=fx⃗(g−1(u,v))⋅∣Jg−1(u,v)∣=e.−uf_{\vec{y}}(u,v)=f_{\vec{x}}(g ^{-1}(u,v)) \cdot \left| J g ^{-1}(u,v) \right| = e ^{.-u}

con

∣Jg−1(u,v)∣=∣det⁡(10−v⋅e−ue−u)∣\left| J g ^{-1}(u,v) \right| = \left| \det\begin{pmatrix} 1 & 0\\ -v \cdot e^{-u} & e^{-u} \end{pmatrix} \right|

Usamos

10,+∞(u)⋅1[0,e−u](v⋅e−u)⋅e−u\mathbb{1}_{0,+\infty}(u) \cdot \mathbb{1}_{[0,e^{-u} ]} (v \cdot e^{-u} ) \cdot e ^{-u} fv⃗(v)=∫−∞+∞1[0,+∞](u)⋅1[0,+∞](v⋅e−u)e−u duf_{\vec{v}} (v) = \int_{-\infty}^{+\infty} \mathbb{1}_{[0,+\infty]}(u)\cdot\mathbb{1}_{[0,+\infty]}(v \cdot e^{-u} ) e^{-u} \, du 0≤v⋅e−u≤e−u  ⟺  0≤v≤10\leq v \cdot e^{-u} \leq e^{-u} \iff_{0}\leq v\leq 1 =1[0,1](v)⋅∫0+∞e−u du=1[0,1](v)=\mathbb{1}_{[0,1]}(v) \cdot \int_{0}^{+\infty} e^{-u} \, du =\mathbb{1}_{[0,1]}(v)

Veamos si se cumple:

fy⃗=1[0,1](v)⋅e−u⋅1[0,+∞](u)f_{\vec{y}}=\mathbb{1}_{[0,1]}(v) \cdot e^{-u} \cdot \mathbb{1}_{[0,+\infty]} (u)

Entonces son independientes.

Temas relacionados