1. Introduction
Let be a population composed of individuals. Each individual is characterized by a d-dimensional vector of the Euclidean space. Each entry of the vector corresponds to a characteristic of the individual. Examples of characteristics can be wealth, height, income, education level, health, age, notoriety, and so on. Thus, any individual “j” may be characterized by a vector with d components, where d is the number of measured characteristics.
If and are two vectors from we say iff for all This is the natural partial order. Let A be a subset of . We shall say that from A is maximal if there is no from A such that . If the set A has an unique maximal element, then it will be called the last element of A.
If are two functions then iff for all
The individuals can be compared with each other using each component. The temptation of making tops is licit at the unidimensional level.
However, in the multidimensional case, the order relation is not total. It follows that the order statistics make no sense. However, people want to rank a set of vectors in order to make a decision. Sometimes it is unavoidable.
In a partially ordered set we shall call the set formed by the first and the last element the extremes set. Of course, the extreme set has 0, 1 or 2 elements.
Definition 1. A partially ordered set S has the property F if it has a first element, that is if there exists an element such that for every . Next, S has the property L (or it has the leadership property) if it has a last element, i.e., if there exists an element such that for every . Sometimes we call the last element “the leader”. Furthermore, finally, S has the property FL if it has both a first element and a last element.
Let be a sequence of d-dimensional random vectors.
Let be a natural number. Consider the sample
Some questions are natural. For instance: which is the probability that a “leader” exists in that sample? is a leader if for any Or which is the probability that an anti-leader does exist in the sample? is an anti-leader if for any . Or which is the probability that in the sample we find both a leader and an anti-leader? Or the probability that the sample contains at least a comparable pair (, ) meaning that or ?
For instance, if are i.i.d. and uniformly distributed in the hypercube then it is easy to see that and the probability to find at least two comparable vectors is in the case (For we do not know the answer!) In the case the probabilities and tend to 0 if . Is that a general rule? Do these probabilities always tend to 0 if no matter of the distribution of the i.i.d. random vectors ?
These questions have prompted our paper. The subject of our paper belongs to the domain of Probability Theory called “Random Geometry”. We recall that Random Geometry covers a variety of techniques and methods applicable for the description of stochastic behavior of geometric objects ranging from graphs and networks to abstract or embedded continuous manifolds. An interesting problem in the Random Geometry is the study of the number of maximal elements of a random set of points in . Let be a set of n random vectors in . Define the random variable as being the number of maximal elements in .
In [
1], Barndorff-Nielsen and Sobel initiated the study of the number of maximal elements of the set
as an attempt to describe the boundary of a set of random points in
. This problem in closely related with a problem from computational geometry, namely the construction of the convex hull of a set [
2]. Potential applications of the study may be found in various disciplines such as mathematical programming, multiple criteria decision making, game theory, algorithm analysis, pattern classification, graphics, economics, data analysis, social psychology, and others.
In [
1], asymptotic results for the expected value and variance of the random variable
were obtained. Subsequent papers that deal with this problem, some of them giving simplified proofs include [
2,
3,
4,
5,
6,
7]. The interested reader can find more information in the books [
8,
9] and in the papers [
5,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19]. Now we shall return to our problem. We searched for answers regarding our specific problem in the literature, but we were not able to find any. This is why we decided to study the probability of naive extrema existence in a sample. The structure of our paper is described below.
In
Section 1 and
Section 2 we introduced definitions, notation, and elementary examples. Some results for the discrete distributions and formulation of three conjectures are presented in
Section 3. They are followed by the analysis of the continuous case, in
Section 4. The last part of the article is dedicated to absolutely continuous distributions. It contains a series of examples that shed light on this subject.
2. Stating of the Problem
Let be a sequence of i.i.d. d-dimensional random vectors defined on a probability space and let F be their common probability distribution. If we shall write only we shall understand that is a copy of
We will denote with the same letter
F both its distribution probability and its distribution function: if
is a Borel set then
but
is a vector from
, then
or, explicitly,
where
. Let also
Let us notice that in the 1-dimensional case, if F is continuous then
In the 2-dimensional case we will prefer the notation Now means . In this particular case the marginals will be denoted by and :
Thus,
is the probability that one of these
n random vectors is the greatest of them,
is the probability that one of them is the smallest and
the probability that the set
has both a minimum and a maximum. Or, in other words,
is the probability that the set
have a leader,
is the probability that the set
have an anti-leader and
is the probability that the set
have both a leader and an anti-leader. It is evident that
Notice that all the probabilities depend only on the distribution F of
When the danger of confusion arises, we shall write instead of to be on the safe side.
In other words, is the probability that the random set has the property L, is the probability that the same set has the property F and is the probability to have the property FL.
Definition 2. We say that
- -
the probability distribution Fhas the leader propertyif ,
- -
the probability distribution Fhas the strong leader propertyif ,
- -
the probability distribution Fhas the anti-leader propertyif ,
- -
the probability distribution Fhas the strong anti-leader propertyif
- -
the probability distribution Fhas the extremes propertyif ,
- -
the probability distribution Fhas the strong extremes propertyif .
We say that the random vector has a leader if its distribution has the leader property. In the same way has the anti-leader property if its distribution has the anti-leader property and so on.
Definition 3. Let be a function.
We say that
- -
is strongly increasing iff and
- -
is strongly decreasingiff
Typical examples of strongly increasing functions are with all increasing. Here are intervals. For instance, for the functions or are strongly increasing for and is strongly decreasing. Here
Let be a dimensional random vector and let F be its probability distribution.
Some statements are obvious:
(S1)If has a leader and is strongly increasing then has a leader, too. Moreover, if has a leader and ϕ is strongly decreasing, then has the anti-leader property.
(S2)The probabilities remain the same if we replace the sequence with where ϕ is a strongly increasing function.
(S3)Let and . If has a leader then has a leader, too. Or, in terms of distributions: if F has the leader property, then all its marginals have the leader property.
(S4)Let σ be a permutation of and If has the leader property, then has the leader property. The probabilities are the same for and
The support of denoted by supp is defined to be the intersection of all closed sets B such that Or, in terms of random vectors: if and only if P for any Of course S is closed. In the unidimensional case—i.e., if —the support is with Moreover, the restriction of the distribution function on S is increasing since if and then
Definition 4. Let . The point is called aweak leaderif there exists no such that In the same way is aweak anti-leaderif there exists no such that After all, a weak leader is a Pareto maximum and a weak anti-leader is a Pareto minimum. Obviously the set has a leaderif it has the property L; that is, if there exists a point such that Sup This leader is unique.
Remark thatthe compact set S has no leaderif and only if it has at least two weak leaders.
Examples. Here the uniform distribution on a set S will be denoted by and will be the Lebesgue measure on
1. with The set S has no leader. All the points of S are weak leaders.
2. where with measurable functions such as
In this case, Supp Here means the closure of If are nondecreasing then S has the leader and the anti-leader
3. where is a compact set such that Let be the th projection of The leader of C is the point provided that it belongs to
3. The Discrete Case
Suppose that the distribution F is discrete: with at most countable, is the Dirac distribution, for all and
In this case, we obtain an exact formula for and also an upper bound.
Proposition 1. Let F be the discrete probability distribution described above.
Consequently if then F has not the leader property.
Proof. According to the definition of we have where
However,
In the same way
and, in general for
The claim is a consequence of the well known formula for the probability of a union of sets. □
Unfortunately, Formula (5) seems too involved in order to be useful.
Proposition 2. Let Z be a discrete dimensional random vector with the distribution function F and let Then the following assertions hold:
a. If the set S has a leader and then the distribution F has the strong leader property. In this case Otherwise written, the occurrence of a leader is almost sure.
b. If S is compact and F has the leader property, then S has the property L. In the same way, if F has the anti-leader property then S has the property F.
c. If S is finite then F has the leader property if and only if S has the property L.
d. If and all are finite, then F has the strong leader property hence
Proof. a. Let be a sequence of i.i.d. F—distributed random vectors. Let and be the leader of As F is discrete it follows that S is at most countable. Let . Let N be the first n such that N is finite and geometrical distributed: Of course hence
b. Suppose that S is compact. We prove that if F has the leader property then S has a leader, too. Suppose the contrary. As S is compact then it has no leader iff it has at least two different weak leaders. Denote them by and Let Surely since no point in S can be greater both than and The complement of S is an open set, hence there exists a neighborhood V of such that Let and be two neighborhoods of and such that if and then By the definition of the support and Let and the hitting times of and and let
If then the set has no leader since nobody is greater both than and if and Thus . As it follows that
c. Apply a. and b.
d. A consequence of a.: is finite and has the leader . Moreover □
Remark 1. The fact that means that all the components of are independent. Thus, the point d. from Proposition 2 is to be read as “If all the components of have finite range and are independent, then has the strong leader property”.
Sometimes one can relax the finitude conditions allowing a component to have an infinity of values. Precisely, it is true that “If all but one components of have finite range and are independent then has the leader property”.
Lemma 1. Let be two independent sequences of distributed i.i.d. random variables. Suppose that their distribution is discrete and moreover, that there exist some positive β such that for all .
Let and Particular case:
Proof. As we can write .
From the well known inequality and the fact that
we infer that
It follows that
However, therefore, by (*) .
On the other side .
Adding these quantities we obtain □
Proposition 3. Let be a sequence of i.i.d. random vectors in the plane. Suppose that , and with and . Suppose moreover that are positive for and and
Then Thus, F has the leader property.
Proof. Let be fixed. Notice that where such that
Consider the random vector whith It is well known that the distribution of is the multinomial one:
and its expected value is
E.
Let Then Consider
Notice that
It follows that Thus
.
However, E.
It results that □
With the same proof one can obtain a small generalization
Corollary 1. Let , be a sequence of i.i.d. - dimensional random vectors, . Suppose that the distribution F of has the form ⊗ whereSupp is a finite set with the property L and such that sup Let be the leader of Supp and Then Thus, F has the leader property.
Conjecture 1. We think that if the support of is finite and has the property L, then has the leader property.
The situation changes if is discrete but is infinite at least for two indices and
Conjecture 2. Let be 2-dimensional vector. If and are infinite then has NOT the leader property.
We were able to prove this guess in two cases. Both conjectures are true if supplementary conditions on and are assumed.
Proposition 4. Let with
If or then hence F has not the leader property.
Proof. Let , be a sequence of i.i.d. F distributed random vectors. Suppose that Let
Then
EE.
With our notation, no matter of
However, According to Lemma 1,
Therefore and the last quantity tends to 0 since , according to Beppo-Levi theorem. □
Remark 2. Let F be a probability distribution on Consider the sequence Proposition 4 says that the conjecture is true if one of the sequences or is bounded. However, the condition fails to be true in important cases such as
In this situation we use another approach.
For any with for all let and be the discrete analog of the hazard rate (see [20], Section 2.2, relation (2.1)). For the continuous case see Definition 5 below, in Section 5. The connection between and is given by for all
As the only condition on the positive numbers is that or, which is the same thing, that
Proposition 5. Let F be a distribution on with the property that Let Then Proof. From the definition of N we see that or or, again
From the elementary inequality which holds for we infer that
For the function attains its maximum at and the maximum is
That is why
As we obtain hence
However, . □
Corollary 2. Suppose that is a probability distribution on with the property that Then F does not have the leader property.
Proof. However,
Notice that It follows that
However, for N to be good enough . It follows that if then too.
The last term obviously vanishes to 0 hence
The same holds for □
Corollary 3. The distribution has never the leader property.
Proof. In this case, and . We shall focus on Now from Proposition 5 becomes
with
It follows that with
Therefore
thus
. In the same way . Therefore, according to Corollary 2, F has not the leader property. □
The problem is that not always is true that . For instance, if the limit is infinity.
Remark 3. Unfortunately, the method fails to answer the problem:
Is it true that has never the leader property when ?
We think the answer is positive.
4. The Continous Case
Recall that a probability distribution F is continuos if for any . It is well known that if F is absolutely continuous with respect to Lebesgue measure then it is continuous.
It is more or less obvious that if
F is continuous the following formulae hold:
All of them are immediate consequences of the continuity of F. Precisely, if are i.i.d. F distributed random vectors then .
Computation rules.
We will need the following result
Lemma 2. Let be a probability space, and be measurable spaces. Let be a random vector. Suppose that the distribution of is decomposable, meaning that it can be written as where is a probability on and the conditioned distribution is a transition probability from to
Let be measurable and bounded. Then
- (i)
E
In the particular case when X and Y are independent, hence
- (ii)
E
Proof. According to the definition of the product between a probability and a transition probability we know that P for every and . It follows that (i) holds for simple functions hence it holds for any bounded measurable ones. □
Proposition 6. Let be a sequence of i.i.d. d-dimensional random vectors. Suppose that is F— distributed. Let Φ: be defined as .
Then the following assertions hold:
(A) = E
(B) Suppose moreover that F is continuous. Thenwhere . In the 2-dimensional case it means that if the distribution of is absolutely continuous and has the density ρ then
.
Proof. This assertions are consequences of Lemma 2. Namely, let be abounded and measurable function and
Let . As and are independent, clearly E
However, are independent too, hence where . Next, as is a product, it follows that
The claims (17) and (18) are simple corollaries of (20):
E
=
E
As about the claim (19): as the sets are disjoint a.s. we have
However, E=E ( E (since are i.i.d.) = E where □
Proposition 7. If the two-dimensional random vector has the leader property and are increasing functions, are intervals, then the random vector has the leader property, too.
Proof. The first assertion is obvious: According to the definition
However, and and
For the second one, let and be the supports of X and As we assumed that X and Y are continuous these two sets are closed and uncountable. Let and , be continuous, increasing such that .
Then iff and iff .
Moreover, esssup esssup □
Proposition 8. Let be a sequence of i.i.d. dimensional continuous random vectors that are F-distributed. Suppose that the support of F, denoted by S, is compact.
If S has not the property L then F has no leader.
Proof. Notice that by the very definition of We prove the 2-dimensional case since the dimensional case has a similar proof.
Let In terms of probability distributions, .
We claim that is a leader if and only if Indeed, if is a leader, then Conversely, if then hence almost surely. It means that is a leader.
We prove that if S has not the property L, then F has no leader. Suppose the contrary: F has a leader, meaning that .
Notice that
For, if , we could find a sequence in S such that As S is compact, this sequence has a convergent subsequence to some and in that case . Recall that F is continuous hence E Then = as , contradicting the hypothesis □
The fact that we have a formula for simplifies a lot the things.
The results from Lemma 1, Corollary 1, Proposition 4 become
Proposition 9. The following assertions hold:
Let F be a continuous probability distribution on . Then
Let with finite and unidimensional continuous distribution. Suppose that has the leader Then
Let with and continuous. Then F has not the leader property.
If and all the distributions are continuous, then
Proof. 1. Let be i.i.d. random variable. We denote by where are i.i.d. distributed random variables. Let S be the support of Notice that X and Y have the same support as and , namely Let
Then ~U are standard uniformly distributed and, what is crucial almost surely due to the fact that is increasing. However, P
2. Let such that that . Denote by Thus, . Let be fixed.
Consider the sets Clearly P. Given the probability to have a leader is from 1. It follows that
3. From proposition 6 we know that and the last integral vanishes as
4. E □
5. The Absolutely Continuous Case
In this section, we study the case when the distribution
F of the i.i.d. sequence of random vectors is absolutely continuous. In
Section 5.1. we establish general conditions under which
F has the leader property. In the following subsections we consider special cases for distribution
F in the general result of
Section 5.1. In
Section 5.2. we investigate the case when
F is the product of an exponential and a convolution of exponential with uniform distribution. In
Section 5.3. we consider the case when the random vectors have an uniform distribution in a triangle and in
Section 5.4. we study the case when
F is a mixture of small uniforms. In
Section 5.5. we have considered the case when
with the exponential distributions
and
In the last subsection we give an example of leaderless distribution.
5.1. A General Result
Here we are looking for finding sufficient conditions in order that F have a leader provided that F is absolutely continuous. The main result of this subsection is Proposition 10.
We will focus on the case The case is much more difficult and remains an open problem.
Thus, the 2-dimensional random vectors become and where Q is a transition probability from to with the meaning that for any Borel set B from .
Or, explicitly, one can write
E for any u bounded and measurable function.
Suppose that is absolutely continuous with respect to Lebesgue measure and let be its density:
Important remark. Whenever possible we have validated our computations by computer simulations. A brief description of the simulation procedure follows: To generate a random vector in plane which is F-distributed we used the usual algorithm: let be the distribution of X and be the distribution of Y given that . Then generate X with the distribution and after that generate Y with the distribution . As long as these distributions are classical the procedure is very fast. We used the language “R”. In our case . The following script in “R”:
x=rexp(n);
u=runif(n);
y=x+u
generates a set of n random vectors which are F-distributed. To estimate we generate N such sets and count the number of cases when has a leader. The proportion of these cases should approximate if N is great. A script which does that could be as follows (the reader can check the script himself)
marex=rep(0,N);
marey=rep(0,N)
for (k in 1:N)
x=rexp(n);
u=runif(n);
y=x+u;
marex[k]=which(x==max(x));
marey[k]=which(y==max(y))
}
an=length(which(marex==marey))/N
Returning to the problem: let n be fixed. Order the random variables as . We consider the random vector This is called the order statistics of .
Let ,
The following facts are well known and can be found in any handbook of order statistics, for instance [
21], Propositions 4.1.2–4.1.4 pp. 183–185 or [
22]:
. The density of is
. The density of is
. The density of is where
. The density of is
. The density of is
Lemma 3. With the above notation the following equalities hold
1. P E
2. PE
As a consequence, if X is not bounded above (meaning that for any real t) then P for any t and if it is not bounded below (meaning that for any then P for any Thus, if X is unbounded both below and above, then and are positive for all
Proof. To prove 1, notice that
However, this is the density of the random variable written traditionally defined by The random variable has the tail Notice that if is positive—interpreted as a life time—the random variable is denoted traditionally by and called the residual time of life at age Its expected value is denoted in demography by e and is denoted by
It follows that Ee no matter of n and
P
Or, in a rigorous writing
P hence
PEEE
EE
2 In the same way
This is the density of the random variable It follows that
P or, explicitly P □
Now we can state a result for non-negative random variables. If they are thought as waiting times, a useful tool is the concept of
hazard rate, also known as
failure rate (see [
23]).
Definition 5. Let be an absolutely continuous random variable, be its tail and its density. The quantity is called the hazard rate of X (or of ). Then If is non-decreasing one says that X is of type IFR (Increasing Failure rate) and if it is non-increasing X is of type DFR (Decreasing Failure Rate). Usually one writes IFR in the first case and DFR in the second one. (A better notation would be, of course, IFR or DFR, but this is the tradition). Of course IFR∩DFR means that X is exponentially distributed. All the random variables of type DFR are not bounded above. (for instance, see [23]). Proposition 10. Let , be a sequence of non-negative i.i.d. random vectors in plane with the distribution Let be the hazard rate of Suppose that
- (i)
for some nonnegative for all and
- (ii)
Then hence F has the leader property.
As a particular case, if DFR then thus
Proof. Let be fixed. According to Lemma 3. (1) P= E
However,
(since ).
In the second case P
As a particular case P
Now suppose that Let and the pairs of and : precisely, iff and Then Indeed, and thus □
The result can be extended to random variables which are unbounded both above and below. We can always write where and Suppose that both and are DFR. Then and as in probability.
Then whenever and if As P tends to infinity, E− E tends to 0.
Corollary 4. If both and are DFR then F has both the leader property and the anti-leader property.
Open question. We were not able to answer the question: If F has both the leader property and the anti-leader property is it true or not that F has the extremes property?
If F has both the leader property and the anti-leader property then F has the extremes property.
5.2. Exponential versus Uniform
As an application of Proposition 10 let us consider the following case:
independent on V. are independent copies of
Of course is both IFR and DFR, and, according to Proposition 10 with , we see that
It follows that has the leader property.
The density of is and the exact limit is
.
After some computations one sees that for
Computer simulations suggest that .
In this very case we can transform the random vector into a bounded one with the same ordering property.
The vector has the support in and its density is where .
Notice that the density of is standard uniform and that is unbounded.
One can prove that the density of is always unbounded if DFR and V is bounded.
A first guess is that if we want F have the leader property and be compact then its density p should be unbounded.
The answer is NO.
5.3. Uniform in a Triangle
Proposition 11. Let and be a random vector uniformly distributed on Let Uniform.
Then F has the leader property if and only if
Proof. The density of is and the density of X is no matter of
Let Let
Let . Notice that is strongly increasing. Let be a sequence of i.i.d. F- distributed random vectors and By (S2), and have the same probabilities .
The densities are
Thus, Exp and Exp with for
There are two cases.
Case 1. According to Proposition 10 with we see that for the vectors the limit of is As and have the same probabilities it follows that hence F has the leader property.
As a particular case, for it follows that
Case 2. We cannot apply Proposition 10. We have to use brute force: namely formulas (17)–(19). We apply them to the original
Change the variable y with As
we obtain
Changing the variable x with
To conclude: if then □
Remark 4. If we substitute the set in the statement of Proposition 11 with the set , or with the set
, then F has the leader property iff .
5.4. Mixture of Small Uniforms
It is instructive to compare the two methods — the probabilistic one and the analytic one — in order to decide if a distribution F has the leader property or not.
Let ...and ...be such Let be a distribution probability on the set of positive integers such that for all
Let , and finally, let be a random vector with the distribution Uniform Uniform Uniform
Its density is 1 and the marginal densities are
1, 1
The probabilistic bound is given by Proposition 10.
Let be defined as:
Let be their inverses. Thus, for all we have
and
and where is the fractional part of s.
Let now Notice that
- -
and, as the mapping is strongly increasing
- -
has the leader property if and only if has the leader property.
Its distribution has the density
and, taking into account the construction of f and g
We see that
The criterion from Proposition 10. ( namely in our case) says that has the leader property if has a bounded hazard rate. Now we estimate the hazard rate: the density of is and its tail is Suppose that Then
. On the same interval , .
It follows that For some upper and lower bounds for are
Lemma 4. Let be an increasing sequence such that Let . Then Lemma
Proof. By Lagrange theorem, we know that
⇒
Apply that to : we obtain the inequality
It follows that for any hence
Thus,
for any □
We arrived at the following
Remark 5. If (or, which is the same, if ) then has the leader property. Precisely,
The analytic approach is to estimate E using brute force:
Notice that if then
E
For the last inequality we used Lemma 4.
If we combine the two approaches we arrive at
Proposition 12. Let and be such that Denote , . Also let be a distribution probability on the set of positive integers such that for all Finally, let be a random vector with the distribution Uniform Uniform Uniform
Now we have a clue to decide ifhas the leader property: if is bounded and X is unbounded. However, what can we say if is unbounded, too?
One is obliged to use Formula (17).
In that case the following result may help.
Lemma 5. The following assertions hold:
Let be continuous function at
Then
Let be increasing and differentiable such that and let f be as above.
Then
Let be continuous.
Then
Proof. 1. Let be arbitrary and . Then Obviously f
where as
2. Let As G is differentiable and increasing we have
hence applying 1) we obtain
3. Let and let , be arbitrary small such that
Let and and write with
and
.
Remark that .
It follows that as .
On the other hand □
An interesting result is
Lemma 6. Let be a F distributed random vector where for some Or, in terms of random variables with V independent of X, uniformly distributed on
Suppose that Supp and is absolutely continuous with respect to Lebesgue measure, Suppose that does exist. Then Proof. Let the density of the random vector Let us mention first the obvious relations:
,
Then the probability verifies the inequality
if and
However, according to Lemma 5.
□
Remark 6. It results that if then F has the leader property. For instance, this happens if or if F is a Pareto distribution. In the last case F has even the strong leader property since .
Here are another two cases when Lemma 5 is applied.
5.5. Exponential versus Exponential
Suppose that
Then the density of
is
and its distribution function is
. After some computation we find
and
The result is
Proposition 13. Suppose that is a positive integer. Then the following assertions hold:
hence has the leader property if
If
The sequence is increasing. Thus,
If then
Proof. 1 Remark that if Thus
(with the substitution ).
The function defined by is increasing
.
Therefore, by assertion (2) of Lemma 5. we have
2 If then
with
Let : thus and Then .
Consequently
and, finally
.
It is well known that
3 Let us put The integral becomes
.
With the new substitution it becomes
.
Now denote
and simplify
: we arrive at
which expands as
Let We claim that the sequence is increasing.
Notice that
It is enough to check that or
which is obvious since the difference between the right hand term and the left one is
(with ). Let . The integral becomes
Let
hence
. We arrive at
with
The function
is injective and
. Changing the integration order we obtain
with
.
As we arrive at .
We do not know but it is easy to see that has the property that .
Indeed, applying the Hospital rule we have .
However, . It follows that . According to Lemma 5. the integral converges to hence
This ends the proof. □
Corollary 5. Let the sets : and : . Consider the distributions on with the densitieswhere λ is a positive integer. Then the following assertions hold:
1F
2F
In the following we shall try to give a justification using computer simulation that our results are correct. We chose the case “exponential vs. exponential” (see proposition 13, 2) since this is one of the rare cases when an analytical expression for can be established.
Therefore, let with and , X and U independent. According to proposition 13, 2 we have: with ln To check that statistically, construct the estimator ((no. of sets which have a Leader)/(no. of simulations)). Then K is distributed as . Chose and accepted risk . The confidence interval of K is between the quantile of 0.001 and the quantile of 0.999. For any n we have made simulations, five times and have obtained the results Here are the results for :
, ;
, ;
, ;
, .
One can easily see that never fall out the confidence interval. The conclusion is that we cannot reject the hypotheses that .
5.6. A Leaderless Distribution
Let
The random vector has the density:
One can easily see that for every fbounded measurable we have
E =E.
Note that the distribution of the random vector is the following:
E (F(Z))= with and
Case 1. Suppose that
If then hence
Next, since we have
=
Let in the second integral The new limits:
Thus,
If we put so becomes or
= and the last integral clearly converges to 0 by Beppo Levi theorem.
Case 2. If we have In this case, we put
. Note that
If we make the substitution we obtain
as .
So in this case, too. We use the following nice fact:
Lemma 7. If n is good enough, then .
Proof. Let Thus, and, integrating by parts,
hence
By Stirling formula
at least for n to be good enough.
Now for n to be good enough. □
Case 3. Now we put
according to the same Lemma.
To conclude:
Proposition 14. The distribution with has never the leader property.
The result is counterintuitive: one would expect that F has the leader property if is small enough. After all, if we have F is comonotone copula, and the last one has even the strong leader, anti-leader properties.