Pourquoi

15

je suppose que

P(A|B)=P(A|B,C)P(C)+P(A|B,¬C)P(¬C)

est correct, alors que

P(A|B)=P(A|B,C)+P(A|B,¬C)

est incorrect.

Cependant, j'ai une "intuition" sur la dernière, c'est-à-dire que vous considérez la probabilité P (A | B) en divisant deux cas (C ou Not C). Pourquoi cette intuition est fausse?

zell
la source
4
Voici un exemple simple pour tester vos équations. Lancez deux pièces indépendantes et équitables. Soit A l'événement où le premier monte la tête, B soit l'événement où le second monte la tête, et C l'événement où les deux arrivent la tête. Est-ce que l'une ou l'autre équation que vous avez écrite est correcte?
A. Rex
4
La loi de la probabilité totale dit que si vous voulez exprimer une probabilité inconditionnelle comme une somme de probabilités conditionnelles, vous devez pondérer par l'événement de conditionnement: par exemple P(A)=P(A|B)P(B)+P(A|B¯)P(B¯)
AdamO

Réponses:

25

Supposons, par exemple , contre facile, que la probabilité de A est 1 , quelle que soit la valeur de C . Ensuite, si nous prenons l' équation incorrecte , nous obtenons:P(A)A1C

P(A|B)=P(A|B,C)+P(A|B,¬C)=1+1=2

Cela ne peut évidemment pas être correct, un ne peut probablement pas être supérieur à . Cela aide à construire l'intuition selon laquelle vous devez attribuer un poids à chacun des deux cas proportionnel à la probabilité de ce cas , ce qui donne la première équation (correcte). .1


Cela vous rapproche de votre première équation, mais les poids ne sont pas tout à fait exacts. Voir le commentaire d'A. Rex pour les poids corrects.

Dennis Soemers
la source
1
Les poids dans la "première (bonne) équation" doivent-ils être et P ( ¬ C ) , ou doivent-ils être P ( C B ) et P ( ¬ C B ) ? P(C)P(¬C)P(CB)P(¬CB)
A. Rex
@ A.Rex C'est un bon point, pour une exactitude totale, je pense que ce devrait être et P ( ¬ C | B ) . Tout (un seul terme) sur le côté gauche de l'équation suppose que B est donné, donc sans hypothèses supplémentaires (comme en supposant que B et C sont indépendants l'un de l'autre), il devrait en être de même à droite à gaucheP(C|B)P(¬C|B)BBC
Dennis Soemers
Imaginez que A | B soit sûr à 200%.
Mark L. Stone
@ MarkL.Stone Est-ce à dire que cela se produit toujours deux fois? ;)
Rétablir Monica
9

La réponse de Dennis a un excellent contre-exemple, réfutant la mauvaise équation. Cette réponse cherche à expliquer pourquoi l'équation suivante est correcte:

P(A|B)=P(A|C,B)P(C|B)+P(A|¬C,B)P(¬C|B).

Comme chaque terme est conditionné à , nous pouvons remplacer tout l'espace de probabilité par B et supprimer le terme B. Cela nous donne:BBB

P(A)=P(A|C)P(C)+P(A|¬C)P(¬C).

Ensuite, vous demandez pourquoi cette équation contient les termes et P ( ¬ C ) .P(C)P(¬C)

The reason is that P(A|C)P(C) is the portion of A in C and P(A|¬C)P(¬C) is the portion of A in ¬C and the two add up to A. See diagram. On the other hand P(A|C) is the proportion of C containing A and P(A|¬C) is the proportion of ¬C containing A - these are proportions of different regions so they don't have common denominators so adding them is meaningless.

pic

Reinstate Monica
la source
2
Not "everything is conditioned on B". In particular, P(C) and P(¬C) are not, so you can't just drop B. Moreover, this might suggest the equation is wrong!
A. Rex
@A.Rex Technically you're right, I should have said every term involving A is conditioned on B (I made a simple substitution A|BA). I will correct the answer.
Reinstate Monica
5
My objection wasn't a technicality. Your diagram correctly proves that P(A)=P(AC)P(C)+P(A¬C)P(¬C), which after conditioning on B becomes P(AB)=P(AB,C)P(CB)+P(AB,¬C)P(¬CB); note that the probabilities of C and ¬C are also conditioned on B. This is not the first equation given in the OP, which is good news, because the first equation given in the OP is not correct.
A. Rex
@A.Rex You are right once again, C must also conditioned on B as the proportion of the probability space contained in C might not be the same as the proportion of B contained in C. This point escaped me. I will revise again.
Reinstate Monica
7

I know you've already received two great answers to your question, but I just wanted to point out how you can turn the idea behind your intuition into the correct equation.

First, remember that P(XY)=P(XY)P(Y) and equivalently P(XY)=P(XY)P(Y).

To avoid making mistakes, we will use the first equation in the previous paragraph to eliminate all conditional probabilities, then keep rewriting expressions involving intersections and unions of events, then use the second equation in the previous paragraph to re-introduce the conditionals at the end. Thus, we start with:

P(AB)=P(AB)P(B)

We will keep rewriting the right-hand side until we get the desired equation.

The casework in your intuition expands the event A into (AC)(A¬C), resulting in

P(AB)=P(((AC)(A¬C))B)P(B)

As with sets, the intersection distributes over the union:

P(AB)=P((ABC)(AB¬C))P(B)

Since the two events being unioned in the numerator are mutually exclusive (since C and ¬C cannot both happen), we can use the sum rule:

P(AB)=P(ABC)P(B)+P(AB¬C)P(B)

We now see that P(AB)=P(ACB)+P(A¬CB); thus, you can use the sum rule on the event on the event of interest (the "left" side of the conditional bar) if you keep the given event (the "right" side) the same. This can be used as a general rule for other equality proofs as well.

P(A(BC))=P(ABC)P(BC)
and similarly for ¬C.

We plug this into our equation for P(AB) as:

P(AB)=P(ABC)P(BC)P(B)+P(AB¬C)P(B¬C)P(B)

Noting that P(BC)P(B)=P(CB) (and similarly for ¬C), we finally get

P(AB)=P(ABC)P(CB)+P(AB¬C)P(¬CB)

Which is the correct equation (albeit with slightly different notation), including the fix A. Rex pointed out.

Note that P(ACB) turned into P(ABC)P(CB). This mirrors the equation P(AC)=P(AC)P(C) by adding the B condition to not only P(AC) and P(AC), but also P(C) as well. I think if you are to use familiar rules on conditioned probabilities, you need to add the condition to all probabilities in the rule. And if there's any doubt whether that idea works for a particular situation, you can always expand out the conditionals to check, as I did for this answer.

YawarRaza7349
la source
2
+1. I think you extracted the equation that OP tried to intuit: P(AB)=P(ACB)+P(A¬CB).
A. Rex
Thanks! That was the main point I wanted to make, but couldn't figure out a high-level explanation why the intersection goes on the left rather than the right, so I used formulas instead. Also, I just noticed you were the one who pointed out the mistake in OP's formula, so I credited you for that. (I probably wouldn't have noticed either, lol.)
YawarRaza7349
2

Probabilities are ratios; the probability of A given B is how often A happens within the space of B. For instance, P(rain|March) is the number of rainy days in March divided by the number of total days in March. When dealing with fractions, it makes sense to split up numerators. For instance,

P(rain or snow|March)=(number of rainy or snowy days in March)(total number of days in March)=(number of rainy days in March)(total number of days in March)+(number of snowy days in March)(total number of days in March)=P(rain|March)+P(snow|March)

This of course assumes that "snow" and "rain" are mutually exclusive. It does not, however, make sense to split up denominators. So if you have P(rain|February or March), that is equal to

(number of rainy days in February and March)(total number of days in February and March).

But that is not equal to

(number of rainy days in February)(total number of days in February)+(number of rainy days in March)(total number of days in March).

If you're having trouble seeing that, you can try out some numbers. Suppose there are 10 rainy days in February and 8 in March. Then we have

(number of rainy days in February and March)(total number of days in February and March)=(10+8)/(28+31)=29.5%

and

(number of rainy days in February)(total number of days in February)+(number of rainy days in March)(total number of days in March)=(10/28)+(8/31)=35.7%+25.8%=61.5%

The first number, 29.5%, is the average of 35.7% and 25.8% (with the second number weighted slightly more because there is are more days in March). When you say P(A|B)=P(A|B,C)+P(A|B,¬C) you're saying that x1+x2y1+y2=x1y1+x2y2, which is false.

Acccumulation
la source
1

If I go to Spain, I can get sunburnt.

P(sunburnt|Spain)=0.2
This tells me nothing about getting sunburnt if not going to Spain, let's say
P(sunburnt|¬Spain)=0.1
This year I'm going to Spain, so
P(sunburnt)=0.2
Letting B=Ω, this is, P(B)=1, your intuition would imply
P(A)=P(A|C)+P(A|¬C)
which by the previous argument, isn't neccesarily true.
sheriff
la source