Ellsberg urn problems with multiple degrees of freedom

Abstract

An Ellsberg urn is filled with n balls of m different colors and each ball is marked in only one color. However, the number of balls of each color is unknown, which should be regarded as uncertain variable. Moreover, since we randomly draw a ball from the urn, Ellsberg urn problems are a mixture of uncertainty and randomness. In order to deal with such problems, we can employ probability theory, uncertainty theory, and chance theory. Until now only the case m = 3 has been studied. This paper is aimed at studying more general cases where m > 3. It is concluded that the chance of drawing a ball of one color is equal to that of drawing a ball of another color.

Keywords

Ellsberg urn uncertainty theory chance theory chance measure

1 Introduction

The urn problem could trace back to the ancient Greek period [18] and biblical times [17]. In 1713, Bernoulli discussed the explicit urn problem of drawing “calculi” from urns. Afterwards, the urn problem refers to the model structured by a known number of urns containing balls of different colors. Why are people interested in urn problems? As explained by Balakrishnan and Kotz [9], an efficient method is given to model “random choice”, and urn model enables people to simulate the random process.

In 1923, Pólya and Eggenberger [2] proposed an urn model (called Pólya urn) containing b black balls and w white balls where b and w are known numbers. Pólya urn model was originally applied to issues dealing with the spread of contagious diseases. In order to deal with Pólya urn problems, probability theory successfully provides a mathematical tool [5]. Furthermore, we can employ stochastic process theory to discuss the problems about repeating the procedure that draw a ball at random and add some balls to the urn [1, 7]. In 1961, Ellsberg [4] presented another urn problems called Ellsberg urn problems: There is an urn containing one hundred balls and those balls are black or yellow but the numbers of black or white balls are unknown. The most remarkable feature of Ellsberg urn problems is that the number of balls in each color is unknown which seems to be connected to the belief degrees of human beings. Many scholars in management and economics studied these problems from a management perspective by the method of voting. For example, Ellsberg [4] showed that people prefer the option with no uncertainty when comparing options with uncertainty against options with no uncertainty. Eliaz and Ortoleva [3] studied Ellsberg urn problems with uncertainty in different dimensions and found that the majority of participants prefer no uncertainty when comparing options with no uncertainty against uncertainty in multiple dimensions.

Can Ellsberg urn problems be modeled by mathematical methods instead of voting? In 2019, Liu [14] provided a rigorous mathematical method to deal with Ellsberg urn problems by probability theory, uncertainty theory and chance theory; we will discuss this in Section 2. Liu discussed Ellsberg urn problem containing 30 red balls and 60 other balls (black or yellow balls) but the numbers of black or yellow balls are unknown. In this problem, there are two colors (black and yellow) and the number of balls in each of them is unknown. Following that, Lio [10] studied Ellsberg urn problems containing n balls which are either black, yellow or red. There are three colors and the number of balls in each of them is unknown. In this paper, we aim at studying Ellsberg urn problems with more than three colors and the number of balls of each color is also unknown. Then, we consider the question: what is the possibility of drawing a ball of each color?

The rest of the paper is organized as follows. Some basic concepts of uncertainty theory and chance theory are introduced in Section 2. In order to reasonably solve the Ellsberg urn problems with more than three colors, we apply probability theory, uncertainty theory and chance theory as rigorous mathematical tools in Section 3. Finally, a brief conclusion will be given in Section 4.

2 Preliminaries

Probability theory [8] is used for modeling frequency generated by samples while uncertainty theory is a mathematical tool for modeling belief degrees evaluated by domain experts. When we don’t have adequate historical samples (indispensable for probability theory), we can invite experts to provide their belief degrees. To model belief degrees, Liu [11] built uncertainty theory as an axiomatic system satisfying normality, duality and subadditivity. Later, the product axiom [12] was added into the axiomatic system.

In some situations, a complex system may contain uncertainty as well as randomness simultaneously. In order to model the phenomena, Liu [15] presented chance theory. Then Liu [15] introduced the fundamental concepts of chance measure and uncertain random variable. For uncertain random variables, there are some important contributions. For example, the operational law was proposed by Liu [16], the law of large numbers was proved by Yao and Gao [20], and the stronger law of large numbers was verified by Shen et al. [19].

2.1 Uncertainty theory

Definition 2.1. (Liu [11]) Let ℒ be a σ-algebra on a nonempty set Γ. A real-valued set function on ℒ is called an uncertain measure if it satisfies the following three axioms:

Axiom 1. (Normality Axiom) M {Γ} =1 for the universal set Γ.

Axiom 2. (Duality Axiom) M {Λ} + M {Λ^c} =1 for any event Λ.

Axiom 3. (Subadditivity Axiom) For every countable sequence of events Λ₁, Λ₂, ⋯ , we have $M {⋃_{i = 1}^{\infty} Λ_{i}} \leq \sum_{i = 1}^{\infty} M {Λ_{i}} .$

The triplet (Γ, ℒ, ℳ) is called an uncertainty space. The product uncertain measure was further defined by Liu [12] as follows:

Axiom 4. (Product Axiom) Let (Γ_k, ℒ _k, ℳ_k) be uncertainty spaces for k = 1, 2, ⋯ . The product uncertain measure ℳ is an uncertain measure satisfying $M {\prod_{k = 1}^{\infty} Λ_{k}} = ⋀_{k = 1}^{\infty} M_{k} {Λ_{k}}$ where Λ_k are arbitrarily chosen events from ℒ_k for k = 1, 2, ⋯, respectively.

An uncertain variable was defined by Liu [11] as a measurable function from an uncertainty space (Γ, ℒ , ℳ) to the set of real numbers such that the set {ξ ∈ B} = {γ ∈ Γ | ξ (γ) ∈ B} is an event for any Borel set B of real numbers. The uncertainty distribution was defined by Liu [11] as $Φ (x) = M {ξ \leq x}, \forall x \in R .$

The uncertain variables ξ₁, ξ₂, ⋯, ξ_n are said to be independent if $M {⋂_{i = 1}^{n} (ξ_{i} \in B_{i})} = ⋀_{i = 1}^{n} M {ξ_{i} \in B_{i}}$ for any Borel sets B₁, B₂, ⋯, B_n of real numbers (Liu [12]).

Suppose that ξ₁, ξ₂, ⋯, ξ_n are independent uncertain variables with regular uncertainty distributions Φ₁, Φ₂, ⋯, Φ_n, respectively, and f (x₁, x₂, ⋯, x_n) are a strictly increasing function with respect to x₁, x₂, ⋯ , x_m and a strictly decreasing function with respect to x_m+1, x_m+2, ⋯ , x_n. Liu [13] proved that the inverse uncertainty distribution of ξ = f (ξ₁, ξ₂, ⋯ , ξ_n) is $\begin{matrix} Φ^{- 1} (α) = f (Φ_{1}^{- 1} ( & α), \dots, Φ_{m}^{- 1} (α), \\ Φ_{m + 1}^{- 1} (1 - α), \dots, Φ_{n}^{- 1} (1 - α)) . \end{matrix}$

To represent the size of uncertain variable ξ, the expected value was defined by Liu [11] as $E [ξ] = \int_{0}^{+ \infty} M {ξ \geq x} d x - \int_{- \infty}^{0} M {ξ \leq x} d x$ provided that at least one of the two integrals is finite. If the inverse uncertainty distribution Φ^-1 (α) of ξ exits, then $E [ξ] = \int_{0}^{1} Φ^{- 1} (α) α .$

2.2 Chance Theory

Let (Γ, ℒ , ℳ) be an uncertainty space, and (Ω, ℱ, Pr) be a probability space. Then (Γ, ℒ , ℳ) × (Ω, ℱ, Pr) is called a chance space. That is $(Γ, ℒ, M) \times (Ω, ℱ, Pr) = (Γ \times Ω, ℒ \times ℱ, M \times Pr)$ where Γ × Ω is the universal set, ℒ × ℱ is the product σ-algebra and ℳ × Pr is the product measure. Liu [15] defined the chance measure of an event Θ in ℒ × ℱ as $\begin{matrix} Ch {Θ} = \\ \int_{0}^{1} Pr {ω \in Ω | M {γ \in Γ | (γ, ω) \in Θ} \geq x} d x . \end{matrix}$ (1) Meanwhile, Liu [15] proved that the chance measure Ch {Θ} satisfies Ch {Γ × Ω} =1 and is a monotone function increasing with respect to Θ. Liu [15] also proved the self-duality of chance measure. That is $Ch {Θ} + Ch {Θ^{c}} = 1$ for any event Θ. Furthermore, Hou [6] showed the subadditivity of chance measure. That is, for every countable sequence of events Θ₁, Θ₂, ⋯, we have $Ch {⋃_{i = 1}^{\infty} Θ_{i}} \leq \sum_{i = 1}^{\infty} Ch {Θ_{i}} .$

An uncertain random variable was defined by Liu [15] as a measurable function from a chance space (Γ, ℒ , ℳ) × (Ω, ℱ, Pr) to the set of real numbers such that the set {ξ ∈ B} is an event in ℒ × ℱ for any Borel set B. For any $x \in R$ , the chance distribution was defined by Liu [15] to describe an uncertain random variable as $Φ (x) = Ch {ξ \leq x} .$ To measure the size of an uncertain random variable ξ, the expected value of ξ was defined by Liu [15] as $E [ξ] = \int_{0}^{+ \infty} Ch {ξ \geq x} d x - \int_{- \infty}^{0} Ch {ξ \leq x} d x$ provided that at least one of the two integrals is finite. If ξ is supposed to have an inverse uncertainty distribution Φ^-1 (α), Liu [15] proved that it has an expected value $E [ξ] = \int_{0}^{1} Φ^{- 1} (α) α .$

3 Ellsberg urn problems

In 1961, Ellsberg [4] proposed an urn problem: red and black balls filled in the urn are in unknown proportion. He then considered these problems from a management perspective by the method of voting. In this section, we study urn problems with multiple colors from a mathematical perspective and consider the fundamental question: what is the possibility of drawing a ball of each color?

In order to describe the general Ellsberg urn problems systematically, we first introduce the concept of degrees of freedom. The degrees of freedom of a system are the number of parameters of the system that may vary independently. Usually, this concept reflects the minimum number of parameters of the system that need to be identified in order to identify all the parameters. Suppose an Ellsberg urn is filled with n balls in m different colors where n and m are positive integers and each ball is marked in only one color. The number of balls of each color is unknown. In this system, the number of balls of each color is a parameter. Totally, there are m parameters. In order to determine all parameters, we need to determine at least (m - 1) parameters. Thus, the degrees of freedom of the system are (m - 1).

In 2019, Liu [14] discussed Ellsberg urn problems with one degree of freedom using probability theory, uncertainty theory, and chance theory. Lio [10] studied the case with two degrees of freedom. In this section, we consider Ellsberg urn problems with more than two degrees of freedom. Suppose an Ellsberg urn is filled with n balls in m different colors where n and m are positive integers and the number of balls of each color is unknown. The first step is numbering. Imagine that all balls are numbered from 1 to n in order of color 1, color 2, ⋯, color m. Take an uncertainty space (Γ, ℒ , ℳ) to be $\begin{matrix} Γ = { & (γ_{1}, γ_{2}, \dots, γ_{m}) | γ_{1} + γ_{2} + \dots + γ_{m} = n, \\ γ_{k} = 0, 1, \dots, n, k = 1, \dots, m} \end{matrix}$ (2) with power set and uncertain measure

$M {Λ} = \frac{| Λ |}{| Γ |} = \frac{| Λ |}{C_{n + m - 1}^{m - 1}}$ (3) where |Λ| represents the cardinality of the event Λ and $C_{n + m - 1}^{m - 1} = \frac{(n + m - 1)!}{n! (m - 1)!} .$ (4)

Example 3.1. Let Γ be the set made up of all possible combinations of differently colored balls in the previously stipulated order. Each element (γ₁, γ₂, ⋯ , γ_m) ∈ Γ means that there are γ_k balls in color k, k = 1, 2, ⋯ , m, respectively.

For example, if m = 1 and n = 1, the urn has one ball in one color resulting in only one case which is in accord with $| Γ | = C_{1}^{0} = 1$ in Equation 4. If m = 2 and n = 2, we have $| Γ | = C_{3}^{1} = 3$ which indicates that the urn with two balls of two colors includes three cases: both balls marked in color 1 or both balls marked in color 2 or one ball marked in color 1 and the other ball marked in color 2.

Since the number of balls of color i is completely unknown for each i with 1 ≤ i ≤ m and can be any integer between 0 and n, Liu [14] suggests treating it as an uncertain variable $ξ_{i} (γ_{1}, γ_{2}, \dots, γ_{m}) = γ_{i} .$ (5) Thus, ξ₁ + ξ₂ + ⋯ + ξ_m = n. In addition, the balls of color i are numbered from $1 + \sum_{j = 1}^{i - 1} γ_{j}$ to $\sum_{j = 1}^{i} γ_{j}$ for i = 1, 2, ⋯ , m, respectively.

Theorem 3.1. Let n and m be positive integers, Γ be a set defined in Equation (2) and ξ₁, ξ₂, ⋯ , ξ_n be uncertain variables defined in Equation (5). For each element (k₁, k₂, ⋯ , k_m) ∈ Γ, we have $M {ξ_{1} = k_{1}, ξ_{2} = k_{2}, \dots, ξ_{m} = k_{m}} = \frac{1}{C_{n + m - 1}^{m - 1}} .$

Proof: It follows from Equation (3) that we have $\begin{matrix} M {ξ_{1} = k_{1}, ξ_{2} = k_{2}, \dots, ξ_{m} = k_{m}} \\ = M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | γ_{1} = k_{1}, γ_{2} = k_{2}, \dots, \\ γ_{m} = k_{m}} \\ = \frac{| {(k_{1}, k_{2}, \dots, k_{m})} |}{| Γ |} \\ = \frac{1}{C_{n + m - 1}^{m - 1}} . \end{matrix}$ The theorem is proved.□

Remark 3.1. Theorem 3.1 suggests that all elements in Γ have the same uncertain measure.

To illustrate the meaning of the representations for the unknown numbers of balls with different colors, we focus attention on the uncertainty distribution of ξ_i for each i with 1 ≤ i ≤ m.

Theorem 3.2. Let n and m be positive integers. For any uncertain variables ξ₁, ξ₂, ⋯ , ξ_m defined in Equation (5), we have $M {ξ_{i} = k} = \frac{C_{n + m - k - 2}^{m - 2}}{C_{n + m - 1}^{m - 1}}$ for k = 0, 1, ⋯ , n, respectively.

Proof: It follows from Equation (3) that $\begin{matrix} M {ξ_{i} = k} \\ = M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | ξ_{i} (γ_{1}, γ_{2}, \dots, γ_{m}) = k} \\ = M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | γ_{i} = k} \\ = \frac{| {γ_{i} = k} |}{| Γ |} \\ = \frac{C_{n + m - k - 2}^{m - 2}}{C_{n + m - 1}^{m - 1}} \end{matrix}$ for i = 1, 2, ⋯ , m and k = 0, 1, ⋯ , n, respectively. The theorem is proved.□

Theorem 3.3. Let n and m be positive integers. The uncertainty distribution of ξ_i defined in Equation 5 is $Φ_{i} (x) = {\begin{matrix} 0, & ifx < 0 \\ 1 - \frac{C_{n + m - k - 2}^{m - 1}}{C_{n + m - 1}^{m - 1}}, & ifk \leq x < k + 1, \\ k = 0, 1, \dots, n - 1 \\ 1, & ifx \geq n \end{matrix}$ for each i with 1 ≤ i ≤ m.

Proof: It follows from Theorem 3.2 that we have $\begin{matrix} Φ_{i} (x) \\ = M {ξ_{i} \leq x} \\ = M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | ξ_{i} (γ_{1}, γ_{2}, \dots, γ_{m}) \leq x} \\ = M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | γ_{i} \leq x} \\ = \frac{| {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | γ_{i} \leq x} |}{| Γ |} \\ = {\begin{matrix} 0, & ifx < 0 \\ \frac{\sum_{j = 0}^{k} C_{n + m - j - 2}^{m - 2}}{C_{n + m - 1}^{m - 1}}, & ifk \leq x < k + 1, \\ k = 0, 1, \dots, n - 1 \\ 1, & ifx \geq n \end{matrix} \\ = {\begin{matrix} 0, & ifx < 0 \\ 1 - \frac{C_{n + m - k - 2}^{m - 1}}{C_{n + m - 1}^{m - 1}}, & ifk \leq x < k + 1, \\ k = 0, 1, \dots, n - 1 \\ 1, & ifx \geq n \end{matrix} \end{matrix}$ for each i with 1 ≤ i ≤ m. The theorem is proved.□

Remark 3.2. It follows from Theorem 3.3 that ξ₁, ξ₂, ⋯, ξ_m are identically distributed.

Theorems 3.1 and 3.3 suggest that it is fair for the balls’ drawing in terms of the definitions of numbering all balls from 1 to n in order of color 1, color 2, ⋯, color m.

The second step is drawing a ball from the Ellsberg urn. Take a probability space (Ω, ℱ, Pr) where Ω = {1, 2, ⋯ , n}, ℱ is the power set of Ω, and probability measure is $Pr {Λ} = \frac{| Λ |}{| Ω |} = \frac{| Λ |}{n} .$ Since the ball is drawn randomly and the unknown number of balls is uncertain, drawing a ball contains uncertainty and randomness. It should be represented by an event in the chance space $(Γ, ℒ, M) \times (Ω, ℱ, Pr) .$

The ball of color 1 is drawn if and only if ω ≤ γ₁ and the ball of color m is drawn if and only if $ω > \sum_{j = 1}^{m - 1} γ_{j} .$ Thus, drawing a ball of color 1 is represented by the event $“ color 1^{″} = {(γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in Γ \times Ω | ω \leq γ_{1}}$ and drawing a ball of color m is the event $" color m " = {(γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in Γ \times Ω | ω > \sum_{j = 1}^{m - 1} γ_{j}} .$

Furthermore, drawing a ball of color i is represented by the event, $" color i " = {(γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in Γ \times Ω | \sum_{j = 1}^{i - 1} γ_{j} < ω \leq \sum_{j = 1}^{i} γ_{j}}$ (6) for each i with 1 ≤ i ≤ m.

Theorem 3.4. Let n and m be positive integers. For each i with 1 ≤ i ≤ m, the chance measure of the event “color i” defined in Equation 6 is $Ch {" color i "} = \frac{1}{m} .$

Proof: When i = 1, it follows from Equation (1) chance measure that the chance measure of drawing a ball of color 1 is $Ch {" color 1 "}$ $\begin{array}{l} = \int_{0}^{1} \Pr {ω \in Ω | ℳ {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | (γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in " color 1 "} \geq x} dx \\ = \int_{0}^{1} \Pr {ω \in Ω | M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | ω \leq γ_{1}} \geq x} dx \\ = \int_{0}^{1} \Pr {ω \in Ω | \frac{C_{n + m - ω - 1}^{m - 1}}{C_{n + m - 1}^{m - 1}} \geq x} dx \\ = \sum_{k=2}^{n+1} \int_{C_{n + m - k - 1}^{m - 1} / C_{n + m - 1}^{m - 1}}^{C_{n + m - k}^{m - 1} / C_{n + m - 1}^{m - 1}} \Pr {ω \in Ω | ω \leq k - 1} dx \\ = \sum_{k=2}^{n+1} \frac{k - 1}{n} \cdot \frac{C_{n + m - k}^{m - 1} - C_{n + m - k - 1}^{m - 1}}{C_{n + m - 1}^{m - 1}} \end{array}$ $\begin{matrix} = \sum_{k = 1}^{n} \frac{C_{n + m - k - 1}^{m - 1}}{n C_{n + m - 1}^{m - 1}} \\ = \frac{1}{m} \sum_{k = 1}^{n} \frac{C_{n + m - k - 1}^{m - 1}}{C_{n + m - 1}^{m}} \\ = \frac{1}{m} . \end{matrix}$ When i = m, the chance measure of drawing a ball of color m is $\begin{array}{l} Ch {" color m "} \\ = \int_{0}^{1} \Pr {ω \in Ω | M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | (γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in " color m "} \geq x} dx \\ = \int_{0}^{1} \Pr {ω \in Ω | M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | ω > \sum_{j = 0}^{m - 1} γ_{1}} \geq x} dx \\ = \int_{0}^{1} \Pr {ω \in Ω | \frac{C_{m + ω - 2}^{m - 1}}{C_{n + m - 1}^{m - 1}} \geq x} dx \\ = \sum_{k=1}^{n} \int_{C_{m + k - 3}^{m - 1} / C_{n + m - 1}^{m - 1}}^{C_{m + k - 2}^{m - 1} / C_{n + m - 1}^{m - 1}} \Pr {ω \in Ω | ω \geq k} dx \\ = \sum_{k=1}^{n} \frac{n - k + 1}{n} \cdot \frac{C_{m + k - 2}^{m - 1} - C_{m + k - 3}^{m - 1}}{C_{n + m - 1}^{m - 1}} \\ = \sum_{k = 1}^{n} \frac{C_{m + k - 2}^{m - 1}}{n C_{m + k - 1}^{m - 1}} \\ = \frac{1}{m} \sum_{k = 1}^{n} \frac{C_{m + k - 2}^{m - 1}}{C_{n + m - 1}^{m}} \\ = \frac{1}{m} \end{array}$ When 2 ≤ i ≤ m - 1, we have $\begin{array}{l} Ch {" color i "} \\ = \int_{0}^{1} \Pr {ω \in Ω | M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | (γ_{1}, γ_{2}, \dots, γ_{m}, ω) \in " color i "} \geq x} dx \end{array}$ $\begin{matrix} = \int_{0}^{1} Pr {ω \in Ω | M {(γ_{1}, γ_{2}, \dots, γ_{m}) \in Γ | \\ \sum_{j = 1}^{i - 1} γ_{j} < ω \leq \sum_{j = 1}^{i} γ_{j}} \geq x} x \\ = \int_{0}^{1} Pr {ω \in Ω | \frac{C_{ω + i - 2}^{i - 1} \cdot C_{n + m - ω - i}^{m - i}}{C_{n + m - 1}^{m - 1}} \geq x} x \\ = \int_{0}^{1} Pr {ω \in {1, 2, \dots, ω^{*}} | \\ \frac{C_{ω + i - 2}^{i - 1} \cdot C_{n + m - ω - i}^{m - i}}{C_{n + m - 1}^{m - 1}} \geq x} x \\ + \int_{0}^{1} Pr {ω \in {ω^{*} + 1, ω^{*} + 2, \dots, n} | \\ \frac{C_{ω + i - 2}^{i - 1} \cdot C_{n + m - ω - i}^{m - i}}{C_{n + m - 1}^{m - 1}} \geq x} x \\ = \sum_{k = 1}^{ω^{*}} \int_{C_{k + i - 3}^{i - 1} \cdot C_{n + m - k - i + 1}^{m - i} / C_{n + m - 1}^{m - 1}}^{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i} / C_{n + m - 1}^{m - 1}} \\ Pr {ω \in Ω | k \leq ω \leq ω^{*}} x \end{matrix}$ $\begin{matrix} + \sum_{k = ω^{*} + 2}^{n + 1} \int_{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i} / C_{n + m - 1}^{m - 1}}^{C_{k + i - 3}^{i - 1} \cdot C_{n + m - k - i + 1}^{m - i} / C_{n + m - 1}^{m - 1}} \\ Pr {ω \in Ω | ω^{*} + 1 \leq ω \leq k - 1} x \\ = \sum_{k = 1}^{ω^{*}} \frac{ω^{*} - k + 1}{n} (\frac{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i}}{C_{n + m - 1}^{m - 1}} - \\ \frac{C_{k + i - 3}^{i - 1} \cdot C_{n + m - k - i + 1}^{m - i}}{C_{n + m - 1}^{m - 1}}) + \sum_{k = ω^{*} + 2}^{n + 1} \frac{k - ω^{*} - 1}{n} \\ (\frac{C_{k + i - 3}^{i - 1} \cdot C_{n + m - k - i + 1}^{m - i}}{C_{n + m - 1}^{m - 1}} - \frac{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i}}{C_{n + m - 1}^{m - 1}}) \\ = \sum_{k = 1}^{n} \frac{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i}}{n C_{n + m - 1}^{m - 1}} \\ = \frac{1}{m} \sum_{k = 1}^{n} \frac{C_{k + i - 2}^{i - 1} \cdot C_{n + m - k - i}^{m - i}}{C_{n + m - 1}^{m}} = \frac{1}{m} \end{matrix}$ where $ω^{*} = ⌈ \frac{(i - 1) n}{k - 1} ⌉$ denotes the greatest integer not exceeding $\frac{(i - 1) n}{k - 1}$ . The theorem thus proved.□

In summary, suppose that an Ellsberg urn contains n balls of m different colors and the number of balls of each color is unknown. The chance measure of drawing a ball of color i is 1/m for each i with 1 ≤ i ≤ m. That means, all events: drawing a ball of color 1, drawing a ball of color 2, ⋯, and drawing a ball of color m have a same chance measure.

4 Conclusion

This paper proposed the general Ellsberg urn problem that contains n balls of m different colors and the number of balls of each color is unknown. The degrees of freedom of this problem are (m - 1). Liu [14] has discussed Ellsberg urn problems with one degree of freedom, and Lio [10] has studied problems with two degrees of freedom. Both of them concluded that the chance measure of drawing a ball of one color is equal to the chance measure of drawing a ball of another color. This paper studied the general cases of more than two degrees of freedom with the help of probability theory, uncertainty theory, and chance theory. It is concluded that drawing a ball of one color has the same chance measure to that drawing a ball in another color. As possible applications, some issues about incomes can be studied in the future.

Footnotes

Acknowledgments

This work was supported by the National Natural Science Foundation of China of No.61873329.

References

Chen

and Wei

, A new urn model, Journal of Applied Probability 4(42) (2005), 964–976.

Eggenberger

and Pólya

, Über die statistik verketteter vorgäge, Zeitschrift für Angewandte Mathematik und Mechanik 3 (1923), 279–289.

Eliaz

and Ortoleva

, Multidimensional Ellsberg, Management Science 62(8) (2016), 2179–2197.

Ellsberg

, Risk, ambiguity, and the Savage axioms, Quarterly Journal of Economics 75 (1961), 643–669.

Friedman

, A simple urn model, Communications on Pure and Applied Mathematics 2(1) (1949), 59–70.

Hou

, Subadditivity of chance measure, Journal of Uncertainty Analysis and Applications 2, Article 14, 2014.

Janardan

, Markov-Polya urn models with predetermined strategies, Gujarat Statistical Review 2(1) (1975), 17–32.

Kolmogorov

, Grundbegriffe der Wahrscheinlichkeitsrechnung, Julius Springer, Berlin, 1933.

Kotz

and Balakrishnan

, Advances in urn models during the past two decades, Advances in Combinatorial Methods and Applications to Probability and Statistics, Birkhäuser, Boston, MA, pp. 203–257, 1997.

10.

Lio

, Two-degree-of-freedom Ellsberg urn problem, Technical Report, 2019.

11.

Liu

, Uncertainty Theory, 2nd edn., Springer-Verlag, Berlin, 2007.

12.

Liu

, Some research problems in uncertainty theory, Journal of Uncertain Systems 3(1) (2009), 3–10.

13.

Liu

, Uncertainty Theory: A Branch of Mathematics for Modeling Human Uncertainty, Springer-Verlag, Berlin, 2010.

14.

Liu

, Uncertain urn problems and Ellsberg experiment, Soft Computing, DOI: 10.n7/s00500-018-3306-7.

15.

Liu

, Uncertain random variables: A mixture of uncertainty and randomness, Soft Computing 17(4) (2013), 625–634.

16.

Liu

, Uncertain random programming with applications, Fuzzy Optimization and Decision Making 12(2) (2013), 153–169.

17.

Rabinovitch

, Studies in the history of probability and statistics XXII probability in the Talmud, Biometrika 56 (1973), 437–441.

18.

Sambursky

, On the possible and probable in ancient Greece, Osiris 12 (1956), 35–481.

19.

Sheng

, Shi

and Qin

, A stronger law of large numbers for uncertain random variables, Soft Computing 22(17) (2018), 5655–5662.

20.

Yao

and Gao

, Law of large numbers for uncertain random variables, IEEE Transactions on Fuzzy Systems 3(24) (2016), 615–621.