Ella, Bella, and Frabduzella are all different heights.
If Ella is taller than Bella, what is the probability that Ella is taller than Frabduzella?
Assume that for any particular height, each girl has the same chance to be that height.
Hint : This conditional probability wiki may prove helpful!
This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try
refreshing the page, (b) enabling javascript if it is disabled on your browser and,
finally, (c)
loading the
non-javascript version of this page
. We're sorry about the hassle.
If you're still thinking the answer is 2 1 , there's some good discussion in the comments to this solution about why this isn't the case.
And with just this information, that probability does not change no matter how many other friends are involved. However, say Stella joins her three friends, and we are told that Ella is also taller than Stella. Now what is probability that Ella is taller than Frabduzella? (I get 3/4.)
Potential follow-up: A second variable, say age, could be added. Say Ella, Bella, Frabduzella and Stella all have different heights and ages. We are told that Ella is taller than Bella and Stella, and is older than Bella. What is the probability that Ella is both taller and older than Frabduzella?
P.S.., Frabduzella?? That's totally original; it didn't get any Google hits at all. :)
Log in to reply
Interesting indeed that even for an infinite number of friends it's the same answer!
Frabduzella... Hahaha! ;)
"However, say Stella joins her three friends, and we are told that Ella is also taller than Stella. Now what is probability that Ella is taller than Frabduzella? (I get 3/4.)"
Indeed. But what if we are instead told that Ella is shorter then Stella? Then the probability that Ella is taller than Frabduzella goes back to being 1/2.
Log in to reply
Yes, I agree. That's interesting; in a way, by bracketing Ella between two other girls in terms of height, we're being told that Ella is of "average" height, in which case Frabduzella is then equally likely to be taller or shorter than Ella.
Well the posted answer is just flat out wrong. The question ask what is the probability that Ella is taller the Frabduzella. There are only two variables in this question. The fact that Ella is taller than Bella has zero relevance to the question at hand. There are only two people being compared and only 1 can be taller than the other. Therefore, the correct answer is 1/2. The 2/3 answer compares all girl, but that is not what the question is asking.
Log in to reply
Without any information there are 6 possible height orderings. The given fact that Ella is taller than Bella eliminates 3 of these orderings, and of the remaining 3 orderings, 2 are such that Ella is taller than Frabduzella, hence the correct answer of 2/3.
Look at an extreme version of this problem. Say there are 100 girls including Ella and Frabduzella, and we are told that Ella is taller than the remaining 98 girls. Informally, this tells us that Ella is a particularly tall girl, and so it would seem more likely than not that she is taller than the remaining girl Frabduzella, and indeed the probability that she is taller is 99/100. In general, If there are n girls including Ella and Frabduzella and we are told than Ella is taller than the other n − 2 girls, then the probability that she is taller than Frabduzella as well is n n − 1 .
Log in to reply
Nice explanation! +1
I'm not a native speaker, so maybe it's a language problem. But doesn't "each girl is equally likely to be any particular height" mean, that there's no gaussian (or any other) distribution of the heights with a maximum height. And wouldn't that mean, that even if Ella is taller than 98 other girls, the 99th girl has a likelyhood of 50% being taller than Ella?
Log in to reply
@Xiao Bao – No.
The expression "each girl is equally likely to be any particular height" is not particularly well-phrased, but the only reasonable way to interpret it is that all orderings of the girls by height are equally likely. This will be true if the heights are IID from a non-atomic distribution, or if the distribution has atoms then if ties are broken randomly.
The problem did not state that all three orders have equal probability. Therefore 1/2 is a likely answer: Ella is either taller or shorter than Frabduzella.
Log in to reply
The problem states that each girl is equally likely to be any particular height, so every possible height ordering, (6 in total, absent any further information), is equally likely. We are then told that Ella is taller than Bella, which eliminates 3 of the orderings, leaving 3 possible orderings, each of which is equally likely to occur.
Log in to reply
Hrmmm... I would have to disagree with your implication
each girl is equally likely to be any particular height, so every possible height ordering, (6 in total, absent any further information), is equally likely
This would be essentially equivalent to "every set of three identically distributed random variables is independent", which is false. In fact independence isn't actually implied by the problem statement.
That all said, under the additional, reasonable assumption the heights are essentially independent, I agree the answer is 2/3.
I do not beleive 2/3 is correct. The 2/3 answer implies a causation that does not exist. I.e. that knowing something about one pair of girls effects the other pair of girls.
It’s like the gambler’s paradox. If you flip a coin and get 99 heads in a row, what is the chance of the 100th flip being heads? 1/2, still. Because the prior 99 flips do not effect the 100th flip.
This is different from the Monty Hall problem. This is because the host does not randomly pick a door to open. He ONLY EVER PICKS A WRONG DOOR. If he randomly picked a door, he would sometimes pick the prize door. But he doesn’t. He always eliminates a wrong choice. In effect, the host causes a change to the initial problem of which door to pick.
Here, that is not the case. The problem states that each girl is just as likely to be any height. The full possibility tree is:
EBF (yes) EFB (yes) BFE (no) BEF (no) FEB (yes) FBE (no)
3/6 or 1/2.
Log in to reply
At first, yes, it is 1/2. However, in light of the new information, 3 of those possibilities can be eliminated, leaving 3 equally likely possibilities. 2 of them result in E > F. Therefore, I believe the probability is 2/3.
Andrew, I suggest you look at Brian Charlesworth's response to Michael Parker (I wish there were a way to link to it!). He provides a very nice explanation of why the information that Ella is taller than Bella affects the (conditional) probability that Ella is taller than Frabduzella.
Note that this is not like the gambler's paradox. In the gambler's paradox the 100th coin flip is an independent coin flip. Here there is no new independent randomness being injected. Rather, a random variable has been sampled (one of six permutations) and then some information about it is revealed (eliminating three of the permutations and leaving three).
To the contrary, this is EXACTLY like the Monty Hall problem. Behind each door is a girl. Choose the door with the largest girl. Then the host opens a door where a girl is who is NOT the largest one. See it now?
Log in to reply
Definitely! And it's similarly to the Monty Hall problem is likely the reason it's stirring up so much controversy...
If I may disagree - the 2/3 answer is based on "equality" of each probability (EBF, EFB, FEB), but this is not granted per se - any girl could be of any possible height; so even knowing that E>B we have not "three equally probable spot to place F" but a continuous line (well, a segment of line) with some random placement of points E and B. And at that segment of all possible length nothing matters except location of E point: judging by that we can find true possibility. And assumption that "each girl is equally likely to be any particular height" is once again showing us to that way of reasoning - any possible "specific" height is equally probable. And we don't have that possibility for an answer. Discarding that, we are left with simple "1/2" if we assume that Ella has medium of all possible heights. If we instead try to find all possible combinations for continuous placements of E and B dots on our segment of "possible heights", we may find another answer, probably, but I think that 2/3 would be unlikely the one.
Log in to reply
Hi Dmitriy... Have you looked at some of the other comments in this problem?
I think that several of them explain quite effectively why the answer should be 2/3 as opposed to 1/2. (e.g. Brian Charlesworth's explanation) Once you have, please let me know if you still have reservations about the answer. Or if you think there is some other way the problem could be stated more clearly...
Even if you use continuous distributions, you will still get the same answer.
What we're doing is essentially an " law of iterated expectation ". Put simply, we condition to the case where the height of the three girls are x > y > z (but not knowing whose height is which), and show that the probability in this restricted outcome scenario is always (say) 2/3, hence the probability in the total outcomes scenario will be this same value.
Of course, if there was some restricted outcome scenario where the probability is a different value, then we will have to do more work (E.g showing that the probability that it happens is 0).
(This digs deeper into the theory, so feel free to tune out here.) However, if there was a non-zero probability that 2 girls could be of the exact same height, then the answer could be different. IE In the case where the heights are x > y = z , the probability is 1, and in the case where the heights are x = y > z , the probability is 0. Try this problem .
This doesn't affect the previous statement because in a continuous distribution, P ( x = y ) = 0 . (Note: we ruled out this possibility in the question by requiring the girls to have different heights)
I disagree. We may discard Bella since the question states they are of all different heights thus there is no correlation nor added information by stating Bellas height. Thus we are left with Ella and Frabduzella. Either Ella is taller or she is not. This is a 50 50 probability.
Log in to reply
I'm not sure that we can disregard Bella as she adds information to the problem... Have you read @Brian Charlesworth 's explanation (in the comments) as to why the probability isn't 1/2? I think he explained it quite effectively!
Log in to reply
@Geoff Pilling This seems to be a common misconception... Perhaps it would help to clarify the solution if you add a pointer to a wiki about conditional probability.
Log in to reply
@Varsha Dani – Good idea... Done!
Here is the link. And I've added it at the bottom of the problem as well.
The question does state they are all different.heights. Thus there is no correlation whatsoever. a is not equal to b is not equal to c. No statement regarding a and b unless all encompassing (which it was not) has a relation to c.
Let me pose it a different way. What is the probability of person A being taller than person B with the assumption that all heights are different? Does this probability change if I then state c, d , and e are taller and f, g , and h are shorter? Or if I state that everyone else in the world is taller? And referencing a previous comment what is the probability of a coin toss if the previous results were 99 tails?
Log in to reply
@Thor Stambaugh "Does this probability change if I then state c, d , and e are taller and f, g , and h are shorter? Or if I state that everyone else in the world is taller?"
Yes, it could. It depends on what new information you get and what question you are asking.
I believe I already replied to the person who brought up the gambler's paradox -- Those coin flips are independent. Here, the heights themselves are independent, it is true. But the comparisons of the heights are not independent.
So, if I have three girls, E B and F, then it is true that comparing any two of them is like a coin flip. But take note of the fact that there are three possible comparisons. E vs B, B vs F, and E vs F. This would be three coin flips, but if they were independent there would be eight possible outcomes . In reality there are only six possible orderings of the heights. Two of the outcomes of the coin flips lead to inconsistent orderings (E >B, B> F, F> E and E<B, B<F, F<E). Thus you cannot make the analogy to the gambler's paradox, and you have to think of this problem in terms of the distribution of all the heights together. Now, being told some information can affect the probability, and in this instance it does.
Note that if you had four girls, E, B, F, and S, then being told that E is taller than B affects the probability that E is taller than F or that B is taller than S, but it does not affect the probability that F is taller than S, which is still 1/2.
Hope this helps.
Log in to reply
@Varsha Dani – Well said, Varsha! :)
Log in to reply
@Geoff Pilling – Yes, this is a definitive explanation. It would almost be worth having staff add this as a note below your solution, so as to prevent the continuing protests that the answer should be 1/2, (although initially these protests did lead to an interesting discussion).
Log in to reply
@Brian Charlesworth – Good call @Brian Charlesworth ... Perhaps @Calvin Lin might have some ideas along these lines?
@Varsha Dani – Thank you Varsha, well stated
Think of it that way...if we just knew that Bella is in the middle but nothing about the other 2...then yes it is absolutely 50-50. But knowing that Ela is taller than Bella gives her an advantage over Frabduzella for being the tallest...
I disagree, the problem already stated that Ella is taller than Bella, leaving the probability that Ella is or isn't taller than Frabduzulla, 50/50.
Log in to reply
@Erick Thompson There has been a fair bit of discussion as to why your assertion is fallacious. Please read the other comments and replies.
It might help to visualize this when we take into account that each girl can have any lenght. So let's say E is 5 meters tall, and B is 4. F can be 1 or 1000. regardless of B. B doesn't give us any extra information. we can add as many extra taller or shorter friends as we want, but they don't matter when we're talking about the probability of A being taller than F. When we say E > B > A > C > D, and ask what the chance is that E is bigger than F, it still doesnt't matter. Let's say that E = 10 meters (it reall doesn't matter, we have an infite scale). So we know 10 > B > A > C > D. F could still be 1000 meters. with the same probability that she would 100000000 meters. or 1 meter. A,B,C and D doen't influence that. Repeat for any value of the variables.
A wonderful discussion problem! I've read all answers and comments and still believe 1/2 to be the correct answer. Although the hint implies that the problem is conditional, the grammar does not. If the question read "If Ella is taller than Bella, what is the probability that Ella is ALSO taller than Frabduzella?", then it would have been a conditional problem.
Without the word ALSO, the question does not appear to be conditional to me and ordering (3! for 6 possibilities) is not relevant to answering the question.
Call me silly, but often asking the question correctly is more difficult than answering it correctly. I think in this case, one word could have easily removed any ambiguity.
Log in to reply
Interesting perspective, Wayne... I know that the wording can be really tricky... In this case, isn't the word ALSO implied by the two statements?
For example, aren't these two synonymous?
Jill is taller than Carl. Jill is taller than Becky.
Jill is taller than Carl. Jill is ALSO taller than Becky.
I'm not sure I understand the difference in meaning... :-/
Log in to reply
I would concur that "also" is redundant from a logical perspective, although its inclusion would rhetorically emphasize the conditionality of the problem. The hint, though, provides enough emphasis in that regard. Another phrasing would mirror P ( E > F ∣ E > B ) , which I read as "Given that Ella is taller than Bella, what is the probability that Ella is taller than Frabduzella?", but I see no reason to change the present wording.
Log in to reply
@Brian Charlesworth – Thanks, Brian. I agree that the hint clarifies what the problem is asking. Drat in that I always try to solve without hints first!
Good reflection, Geoff. I agree that the second sentences are synonymous in your examples.
The difference is that with your first set of sentences you could have said the second fact without having said the first. (Just "Jill is taller than Becky.") - And I would have thought ... OK. Cool, and I don't care about Carl (not relevant). Yet In the second set of sentences, you could not just say "Jill is ALSO taller than Becky." without having first gotten a fact about a third individual (forcing me into a three person comparison ... It could be Jill is taller than Carl, or Carl is taller than Becky, or even Jill is taller than Wayne).
The word "also" here demands that you encompass at least one additional fact, where without the word "also" it is not necessary.
I could twist that a little as it relates to our problem: What if I made a statement and then a question from your example in my discussion of Carl, Jill and Becky...
Jill is taller than Carl. What is the probability that Jill is taller than Becky? - Is Carl relevant here? IMHO, no.
Jill is taller than Carl. What is the probability that Jill is also taller than Becky? - Carl must be relevant here.
Also ... have a good night !
The problem, as worded, does not make the relative height between Ella And Frabduzella conditional upon the relative height of Ella and Bella. As stated, Ella is either taller than Frabduzella or shorter than Frabduzella. The relative height of Bella is irrelevant to that condition, since that relationship is already established.
This is similar to the Monti Hall problem
Log in to reply
Definitely!
its like conditional probability. like socks in a drawer 11 black socks 27 white socks. the probability to pull 2 black socks are 11/38 for one sock then 10/37 for the second black sock. giving 11/38×10/37= 55/703. or there are 55 pairs from 11 or 11Cr2 reads 11 choose 2 and 38Cr2 ....38 choose 2 =703 so 55/703. 55pairs of black over 703 total pairs
Log in to reply
Except the comparison of E to B has nothing to do with the comparison of E to F. They are all different heights: yes. They are equally likely to be any height : Yes. There is no other causal relationship to link B to F, so the probability is 50%
Log in to reply
The claim that "They are equally likely to be any height" is not true once we add the prior information that "E is greater than F".
As an explicit calculation, suppose that
E
,
B
,
F
are uniformly distributed on
[
0
,
1
]
.
Then, simply knowing that
E
>
B
would restrict the probability space to:
The prior distribution of E is uniform, but the posterior distribution of E is not uniformly distributed on [ 0 , 1 ] . You can show that the cdf of E is x 2 and so the pdf of E is 2 x .
For the 0.5 folks, to paraphrase Blitzstein’s discussion of Monty Hall, suppose there were 100 girls with heights h0, h1, ..., h99, all from the same distribution. The probability h0 > h1 = 0.5. However, you are now told h0 > hi for i = 2...99. (98 other doors are open, all with goats behind them...) Do you still think P(h0 > h1) = 0.5 given this additional information? Formally, P(h0 > h1) = 0.5, but P(h0 > h1 | h0 > hi, i > 2) is clearly much larger.
The three friends are Ella ( E ) , Bella, ( B ) , Frabduzella ( F ) . Total there are three are three possibilities keeping one condition constant : E > B ⎩ ⎪ ⎨ ⎪ ⎧ E > B > F ( √ ) E > F > B ( √ ) F > E > B ( × )
∴ The probability that that Ella is taller than Frabduzella is 3 2 .
How do you make a tick or cross in Latex? Or are they square roots and times?
Log in to reply
I didn't use any latex for them. I kept this answer using mobile and there will be default tick and cross mark in the keyboard.
Log in to reply
Ok, thanks. Though it looks like \sqrt and \times. That's fine, thanks for replying
I sort of made an educated guess. I thought that since Ella was taller than Bella, on average Ella will have a height greater than the median. So on average Ella will be taller than Frabduzella. So that left 2/3 and 3/4 as they are both greater than a half. I guessed 2/3 as it seemed more reasonable. I really need to brush up on some probability/statistics :)
Your instinct for probability seems just fine. If you had realized that there are 6 ways to shuffle three things (3!) you could have scrapped 3/4, and your instincts would have led you all the way.
3 out of these 6 orders can be scrapped because Ella is taller than Bella (and the opposite is the case in the other half). Of the remaining orders, there are 2 with Ella the tallest (Bella > Frabduzella, and Frabduzella < Bella).
Log in to reply
Thanks Roland. I think I understand the other solutions presented. And I like your idea of scrapping 3/4 using the 6 ways of shuffling 3 things.
The truth is that at 71 I have forgotten much. I think your approach shows thought, but still a guess...but educated.
But it is given that Ella is taller than Bella. Then why should we consider it in the probability?
Log in to reply
It's given, so I looked at what that meant about the relative height of Ella compared to all possible heights. My logic was that since "for any particular height, each girl has the same chance to be that height" and Ella is taller than Bella, Ella is probably a relatively tall girl. So Ella is probably taller than Frabduzella, hence I eliminated the probabilities less than or equal to a half, because with those probabilities on average Ella is shorter than or the same height as Frabduzella.
Brian Charlesworth's example seems helpful: "Look at an extreme version of this problem. Say there are 100 girls including Ella and Frabduzella, and we are told that Ella is taller than the remaining 98 girls. Informally, this tells us that Ella is a particularly tall girl, and so it would seem more likely than not that she is taller than the remaining girl Frabduzella, and indeed the probability that she is taller is 99/100."
And my approach is definitely more informal, and more intuition based.
Good old Monty Hall paradox ^^
Nice write up, Michael!
Thanks. :D (Very late response, I know.)
This a conditional probability problem. Our initial possible events set is:
B < E < F
B < F < E
E < B < F
E < F < B
F < B < E
F < E < B
We have the information that Ella is taller than Bella, so we can remove all possibilities where Bella is taller than Ella. Now we have:
B < E < F
B < F < E
F < B < E
In only 2 of these 3 cases Ella is taller than Frabduzella. So the probability must be 3 2
I solved the problem like this :-)
For anyone that's interested, this is a variation on a classic statistical problem called the "Monty Hall Dilemma". Famously, it featured on an American game show, and the "smartest woman in the world", Marilyn vos Savant, was the first one to solve it correctly.
Hahaha... Yes! And many Brilliant minds weren't convinced at the time... ;-)
If this is a Monty Hall Problem, the answer should be 1/2. Because it has been shown that one of the possibilitys, (that one of the women are shorter) is no longer possible. This means that the probability at that point is has been reduced from 1/3 to 1/2. This is the point of the Monty Hall.
Log in to reply
I'm not so sure... Doesn't the Monty Hall Problem suggest that the answer should be 2/3?
To start with, the probability for any of the girls being tallest is 1/3.
Knowing E is taller than B adds 1/3 chance for E being tallest, making it 2/3, while F remains at 1/3.
I was first thinking about the odds that Ella would be taller than both Bella and Frabduzella, not caring about that I already have half that information.
The odds that Ella is tallest is 3 1 .
The odds that Ella is taller than Bella is 2 1 . The odds that Ella is taller than Frabduzella is 2 1 . That mean that the odds that Ella is taller than both Bella and Frabduzella is 2 1 ∗ 2 1 = 4 1 which I clearly see is wrong otherwise the odds that any girl would be tallest would be 4 3 instead of 1.
This mean that the odds that one girl is taller than both the other girls are less than the odds that she individually is taller than both girls, one at a time. Since we now that the odds that Ella is taller than Bella is 2 1 we call the odds that she is taller than Frabduzella x. This lead to that the odds that Ella is taller than Bella ( 2 1 ) multiplied with the odds that Ella is taller than Frabduzella ( x ) should be 3 1 .
That gave me 2 1 ∗ x = 3 1 ⟹ x = 3 2
Same of Monty Hall problem
At the start each has equal probability of being the tallest of 1/3 each. Now two girls are decided so it's now 2/3 vs 1/3
Problem Loading...
Note Loading...
Set Loading...
The order of the heights can be
each with equal probability.
Therefore, the chance that Ella is taller than Frabduzella is 3 2
For those of you still convinced that 1/2 must be the right answer, I think @Varsha Dani explains it best in the comments below, copied here:
================================
"Does this probability change if I then state c, d , and e are taller and f, g , and h are shorter? Or if I state that everyone else in the world is taller?"
Yes, it could. It depends on what new information you get and what question you are asking.
I believe I already replied to the person who brought up the gambler's paradox -- Those coin flips are independent. Here, the heights themselves are independent, it is true. But the comparisons of the heights are not independent.
So, if I have three girls, E B and F, then it is true that comparing any two of them is like a coin flip. But take note of the fact that there are three possible comparisons. E vs B, B vs F, and E vs F. This would be three coin flips, but if they were independent there would be eight possible outcomes. In reality there are only six possible orderings of the heights. Two of the outcomes of the coin flips lead to inconsistent orderings (E >B, B> F, F> E and E<B, B<F, F<E). Thus you cannot make the analogy to the gambler's paradox, and you have to think of this problem in terms of the distribution of all the heights together. Now, being told some information can affect the probability, and in this instance it does.
Note that if you had four girls, E, B, F, and S, then being told that E is taller than B affects the probability that E is taller than F or that B is taller than S, but it does not affect the probability that F is taller than S, which is still 1/2.
Hope this helps.