Il teorema di Bayes regge le aspettative?

18

È vero che per due variabili casuali $A$ e $B$ ,

E (A ∣ B) = E (B ∣ A) \frac{E (A)}{E (B)} ?

$E(A\mid B)=E(B\mid A)\frac{E(A)}{E(B)}?$

bayesian mathematical-statistics

— Tomka
fonte

3

Hmm ... Non penso che queste due parti siano equivalenti

— Jon

6

Come sottolineato nelle risposte, la domanda è probabilisticamente insignificante a causa dell'integrazione di variabili casuali da un lato che sono le variabili condizionanti dall'altro lato.

— Xi'an,

25

\begin{matrix} (1) & E [A ∣ B] \overset{?}{=} E [B ∣ A] \frac{E [A]}{E [B]} \end{matrix}

$E[A\mid B] \stackrel{?}= E[B\mid A]\frac{E[A]}{E[B]} \tag 1$ Il risultato congetturato

(1)

$(1)$ è banalmente vero pervariabili casualiindipendenti

A

$A$ e

B

$B$ con mezzi diversi da zero.

Se $E[B]=0$ , il lato destro di $(1)$ comporta una divisione per $0$ e quindi $(1)$ senso. Notare che $A$ e $B$ sono indipendenti o no non è rilevante.

In generale , non vale per variabili casuali dipendenti ma si possono trovare esempi specifici di e soddisfacenti soddisfacenti . Nota che dobbiamo continuare a insistere sul fatto che , altrimenti il lato destro di senso. Tieni presente che è una variabile casuale che sembra essere una funzione della variabile casuale , ad esempio $(1)$ $A$ $B$ $(1)$ $E[B]\neq 0$ $(1)$ $E[A\mid B]$ $B$ mentre è unavariabile casualeche è unafunzionedella variabile casuale , diciamo . Quindi, è simile a chiedere se $g(B)$ $E[B\mid A]$ $A$ $h(A)$ $(1)$

può essere una vera affermazione, e ovviamente la risposta è chenon può essere un multiplo diin generale.

\begin{matrix} (2) & g (B) \overset{?}{=} h (A) \frac{E [A]}{E [B]} \end{matrix}

$g(B)\stackrel{?}= h(A)\frac{E[A]}{E[B]} \tag 2$

g (B)

$g(B)$

h (A)

$h(A)$

Per quanto ne so , ci sono solo due casi speciali in cui può contenere. $(1)$

Come notato sopra, per le variabili casuali indipendenti e , e sono variabili casuali degenerate (chiamate costanti da persone statisticamente analfabete) che equivalgono rispettivamente a ed , e quindi se , abbiamo uguaglianza in . $A$ $B$ $g(B)$ $h(A)$ $E[A]$ $E[B]$ $E[B]\neq 0$ $(1)$
All'altra estremità dello spettro dall'indipendenza, supponiamo che dove sia una funzione invertibile e quindi e siano variabili casuali totalmente dipendenti. In questo caso, $A=g(B)$ $g(\cdot)$ $A=g(B)$ $B=g^{-1}(A)$ e così diventa
$E [A ∣ B] = g (B), E [B ∣ A] = g^{- 1} (A) = g^{- 1} (g (B)) = B$ $E[A\mid B] = g(B), \quad E[B\mid A] = g^{-1}(A) = g^{-1}(g(B)) = B$ $(1)$ che vale esattamente quandodovepuò essere qualsiasi numero reale diverso da zero. Pertanto,vale ogni volta cheè un multiplo scalare di, e ovviamentedeve essere diverso da zero (cfr.Risposta di Michael Hardy). Lo sviluppo sopra mostra chedeve essere unafunzionelinearee che non puòvalereperaffine $g (B) \overset{?}{=} B \frac{E [A]}{E [B]}$ $g(B)\stackrel{?}= B\frac{E[A]}{E[B]}$ $g(x) = \alpha x$ $\alpha$ $(1)$ $A$ $B$ $E[B]$ $g(x)$ $(1)$ functions $g(x) = \alpha x + \beta$ with $\beta \neq 0$ . However, note that Alecos Papadopolous in his answer and his comments thereafter claims that if $B$ is a normal random variable with nonzero mean, then for specific values of $\alpha$ and $\beta\neq 0$ that he provides, $A=\alpha B+\beta$ and $B$ satisfy $(1)$ . In my opinion, his example is incorrect.

In un commento su questa risposta, Huber ha suggerito di considerare l'uguaglianza congetturata simmetrica che naturalmente tiene sempre per le variabili casuali indipendenti indipendentemente dai valori di e e per multipli scalari anche. Naturalmente, più banalmente, vale per

\begin{matrix} (3) & E [A ∣ B] E [B] \overset{?}{=} E [B ∣ A] E [A] \end{matrix}

$E[A\mid B]E[B] \stackrel{?}=E[B\mid A]E[A]\tag{3}$

E [A]

$E[A]$

E [B]

$E[B]$

A = α B

$A = \alpha B$

(3)

$(3)$ any zero-mean random variables

A

$A$ and

B

$B$ (independent or dependent, scalar multiple or not; it does not matter!):

E [A] = E [B] = 0

$E[A]=E[B]=0$ is sufficient for equality in

(3)

$(3)$ . Thus,

(3)

$(3)$ might not be as interesting as

(1)

$(1)$ as a topic for discussion.

— Dilip Sarwate
fonte

9

+1. To be generous, the question could be interpreted as asking whether

E (A | B) E (B) = E (B | A) E (A)

$E(A|B)E(B)=E(B|A)E(A)$ , where the question of division by zero disappears.

— whuber

1

@whuber Thanks. My edit addresses the more general question as to whether it is possible to have

E [A ∣ B] E [B] = E [B ∣ A] E [A]

$E[A\mid B]E[B]=E[B\mid A]E[A]$ .

— Dilip Sarwate

11

The result is untrue in general, let us see that in a simple example. Let $X \mid P=p$ have a binomial distribution with parameters $n,p$ and $P$ have the beta distrubution with parameters $(\alpha, \beta)$ , that is, a bayesian model with conjugate prior. Now just calculate the two sides of your formula, the left hand side is $\DeclareMathOperator{\E}{\mathbb{E}} \E X \mid P = nP$ , while the right hand side is

E (P ∣ X) \frac{E X}{E P} = \frac{α + X}{n + α + β} \frac{α / (α + β)}{n α / (α + β)}

$\E( P\mid X) \frac{\E X}{\E P} = \frac{\alpha+X}{n+\alpha+\beta} \frac{\alpha/(\alpha+\beta)}{n\alpha/(\alpha+\beta)}$ and those are certainly not equal.

— kjetil b halvorsen
fonte

2

The conditional expected value of a random variable $A$ given the event that $B=b$ is a number that depends on what number $b$ is. So call it $h(b).$ Then the conditional expected value $\operatorname{E}(A\mid B)$ is $h(B),$ a random variable whose value is completely determined by the value of the random variable $B$ . Thus $\operatorname{E}(A\mid B)$ is a function of $B$ and $\operatorname{E}(B\mid A)$ is a function of $A$ .

The quotient $\operatorname{E}(A)/\operatorname{E}(B)$ is just a number.

So one side of your proposed equality is determined by $A$ and the other by $B$ , so they cannot generally be equal.

(Perhaps I should add that they can be equal in the trivial case when the values of $A$ and $B$ determine each other, as when for example, $A = \alpha B, \alpha \neq 0$ and $E[B]\neq 0$ , when

E [A ∣ B] = α B = E [B ∣ A] \cdot α = E [B ∣ A] \frac{α E [B]}{E [B]} = E [B ∣ A] \frac{E [A]}{E [B]} .

$E[A\mid B] = \alpha B = E[B\mid A]\cdot\alpha = E[B\mid A]\frac{\alpha E[B]}{E[B]} = E[B\mid A]\frac{E[A]}{E[B]}.$ But functions equal to each other only at a few points are not equal.)

— Michael Hardy
fonte

You mean they are not necessarily equal? I mean they CAN be equal?

— BCLC

1

@BCLC : They are equal only in trivial cases. And two functions equal to each other at some points and not at others are not equal.

— Michael Hardy

2

"But only in that trivial case can they be equal" (emphasis added) is not quite correct. Consider independent

A

$A$ and

B

$B$ with

E [B] \neq 0

$E[B]\neq 0$ . Then,

E [A ∣ B] = E [A]

$E[A\mid B] = E[A]$ while

E [B ∣ A] = E [B]

$E[B\mid A] = E[B]$ and so

E [B ∣ A] \frac{E [A]}{E [B]} = E [B] \frac{E [A]}{E [B]} = E [A] = E [A ∣ B] .

$E[B\mid A] \frac{E[A]}{E[B]} = E[B]\frac{E[A]}{E[B]} = E[A] = E[A\mid B].$

— Dilip Sarwate

@DilipSarwate I was about to say that haha!

— BCLC

I edited your answer to add a few details for the case you pointed out. Please roll back if you don't like the changes.

— Dilip Sarwate

-1

The expression certainly does not hold in general. For the fun of it, I show below that if $A$ and $B$ follow jointly a bivariate normal distribution, and have non-zero means, the result will hold if the two variables are linear functions of each other and have the same coefficient of variation (the ratio of standard deviation over mean) in absolute terms.

For jointly normals we have

E (A ∣ B) = μ_{A} + ρ \frac{σ_{A}}{σ_{B}} (B - μ_{B})

$\operatorname{E}(A \mid B) = \mu_A + \rho \frac{\sigma_A}{\sigma_B}(B - \mu_B)$

and we want to impose

μ_{A} + ρ \frac{σ_{A}}{σ_{B}} (B - μ_{B}) = [μ_{B} + ρ \frac{σ_{B}}{σ_{A}} (A - μ_{A})] \frac{μ_{A}}{μ_{B}}

$\mu_A + \rho \frac{\sigma_A}{\sigma_B}(B - \mu_B) = \left[\mu_B + \rho \frac{\sigma_B}{\sigma_A}(A - \mu_A)\right]\frac{\mu_A}{\mu_B}$

⟹ μ_{A} + ρ \frac{σ_{A}}{σ_{B}} (B - μ_{B}) = μ_{A} + ρ \frac{σ_{B}}{σ_{A}} \frac{μ_{A}}{μ_{B}} (A - μ_{A})

$\implies \mu_A + \rho \frac{\sigma_A}{\sigma_B}(B - \mu_B) = \mu_A + \rho \frac{\sigma_B}{\sigma_A}\frac{\mu_A}{\mu_B}(A - \mu_A)$

Simplify $\mu_A$ and then $\rho$ , and re-arrange to get

B = μ_{B} + \frac{σ_{B}^{2}}{σ_{A}^{2}} \frac{μ_{A}}{μ_{B}} (A - μ_{A})

$B = \mu_B +\frac{\sigma^2_B}{\sigma^2_A}\frac{\mu_A}{\mu_B}(A - \mu_A)$

So this is the linear relationship that must hold between the two variables (so they are certainly dependent, with correlation coefficient equal to unity in absolute terms) in order to get the desired equality. What it implies?

First, it must also be satisfied that

E (B) \equiv μ_{B} = μ_{B} + \frac{σ_{B}^{2}}{σ_{A}^{2}} \frac{μ_{A}}{μ_{B}} (E (A) - μ_{A}) ⟹ μ_{B} = μ_{B}

$E(B) \equiv \mu_B = \mu_B+\frac{\sigma^2_B}{\sigma^2_A}\frac{\mu_A}{\mu_B}(E(A) - \mu_A) \implies \mu_B = \mu_B$

so no other restirction is imposed on the mean of $B$ ( or of $A$ ) except of them being non-zero. Also a relation for the variance must be satisfied,

Var (B) \equiv σ_{B}^{2} = {(\frac{σ_{B}^{2}}{σ_{A}^{2}} \frac{μ_{A}}{μ_{B}})}^{2} Var (A)

$\operatorname{Var}(B) \equiv \sigma^2_B = \left(\frac{\sigma^2_B}{\sigma^2_A}\frac{\mu_A}{\mu_B}\right)^2\operatorname{Var}(A)$

⟹ {(σ_{A}^{2})}^{2} σ_{B}^{2} = {(σ_{B}^{2})}^{2} σ_{A}^{2} {(\frac{μ_{A}}{μ_{B}})}^{2}

$\implies \left(\sigma^2_A\right)^2\sigma^2_B = \left(\sigma^2_B\right)^2\sigma^2_A\left(\frac{\mu_A}{\mu_B}\right)^2$

⟹ {(\frac{σ_{A}}{μ_{A}})}^{2} = {(\frac{σ_{B}}{μ_{B}})}^{2} ⟹ ({cv}_{A})^{2} = ({cv}_{B})^{2}

$\implies \left(\frac{\sigma_A}{\mu_A}\right)^2 = \left(\frac{\sigma_B}{\mu_B}\right)^2 \implies (\text{cv}_A)^2 = (\text{cv}_B)^2$

⟹ | {cv}_{A} | = | {cv}_{B} |

$\implies |\text{cv}_A| = |\text{cv}_B|$

which was to be shown.

Note that equality of the coefficient of variation in absolute terms, allows the variables to have different variances, and also, one to have positive mean and the other negative.

— Alecos Papadopoulos
fonte

1

Isn't this a convoluted way to

A = α B

$A = \alpha B$ where

α

$\alpha$ is some scalar?

— Matthew Gunn

1

@MatthewGunn Your comment is right on target. Normality has nothing to do with the matter. For random variables

A

$A$ and

B

$B$ such that

A = α B

$A = \alpha B$ ,

E [A ∣ B] = α B = A

$E[A\mid B] = \alpha B = A$ and similarly,

E [B ∣ A] = B

$E[B\mid A] = B$ . Consequently, assuming that

E [B] \neq 0

$E[B]\neq 0$ ,

E [A ∣ B] = α B = E [B ∣ A] \cdot α = E [B ∣ A] \frac{α E [B]}{E [B]} = E [B ∣ A] \frac{E [A]}{E [B]} .

$E[A\mid B] = \alpha B = E[B\mid A]\cdot\alpha = E[B\mid A]\frac{\alpha E[B]}{E[B]} = E[B\mid A]\frac{E[A]}{E[B]}.$ No normality, no

| c v_{A} | = | c v_{B} |

$|cv_A|=|cv_B|$ etc, and actually just a rehash of a comment in Michael Hardy's answer.

— Dilip Sarwate

If you write \text{Var} instaed of \operatorname{Var} then you'll see

a Var X

$a\text{Var}X$ and

a Var (X)

$a\text{Var}(X)$ instead of

a Var X

$a\operatorname{Var}X$ and

a Var (X) .

$a\operatorname{Var}(X).$ That's why the latter is standard usage.

— Michael Hardy

@MatthewGun It seems to me that providing answers that contain specific examples is considered valuable content in this site. So yes, when a random variable is an affine function of another, and they are jointly normal with non-zero means, then one needs to have equal coefficients of variation, while, also there are no restrictions on the means of these rv's. On the other hand, when a random variable is just a linear function of another, the relation holds always. So no my answer is not a convoluted way to say

A = a B

$A=aB$ . (cc:@DilipSarwate)

— Alecos Papadopoulos

2

If

B

$B$ is a non-normal random variable with

E [B] = μ_{B} \neq 0

$E[B]=\mu_B\neq 0$ and

A = c B + d

$A=c B+d$ (and so

B = \frac{A - d}{c}

$B=\frac{A-d}{c}$ ), then

E [A ∣ B] = c B + d = A, E [B ∣ A] = \frac{A - d}{c} = B .

$E[A\mid B]=cB+d=A, E[B\mid A]=\frac{A-d}{c}=B.$ Now, if we want to have

E [A ∣ B] = c B + d

$E[A\mid B]=cB+d$ to equal

E [B ∣ A] \cdot \frac{μ_{A}}{μ_{B}} = B \cdot \frac{μ_{A}}{μ_{B}}

$E[B\mid A]\cdot\frac{\mu_A}{\mu_B} =B\cdot\frac{\mu_A}{\mu_B}$ , it must be that

c B + d = B \cdot \frac{μ_{A}}{μ_{B}} ⟹ d = 0, c = \frac{μ_{A}}{μ_{B}}

$cB+d=B\cdot\frac{\mu_A}{\mu_B}\implies d=0,c=\frac{\mu_A}{\mu_B}$ and so

A = c B = \frac{μ_{A}}{μ_{B}} B

$A=cB=\frac{\mu_A}{\mu_B}B$ . So, for nonnormal

B

$B$ , the OP's conjectured result holds if

A = c B

$A=cB$ but not if

A = c B + d, d \neq 0

$A=cB+d, d\neq 0$ .Of course, as you have proved, the result holds for normal random variables if

A = c B + d, d \neq 0

$A=cB+d, d\neq 0$ .

— Dilip Sarwate