Next Article in Journal
A New Extragradient-Viscosity Method for a Variational Inequality, an Equilibrium Problem, and a Fixed Point Problem
Previous Article in Journal
Mechanical and Civil Engineering Optimization with a Very Simple Hybrid Grey Wolf—JAYA Metaheuristic Optimizer
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Stability of Cauchy–Stieltjes Kernel Families by Free and Boolean Convolutions Product

by
Ayed. R. A. Alanzi
1,
Shokrya S. Alshqaq
2 and
Raouf Fakhfakh
1,*
1
Department of Mathematics, College of Science, Jouf University, Sakaka P.O. Box 2014, Saudi Arabia
2
Department of Mathematics, College of Science, Jazan University, P.O. Box 114, Jazan 45142, Saudi Arabia
*
Author to whom correspondence should be addressed.
Mathematics 2024, 12(22), 3465; https://doi.org/10.3390/math12223465
Submission received: 7 October 2024 / Revised: 4 November 2024 / Accepted: 5 November 2024 / Published: 6 November 2024
(This article belongs to the Section D1: Probability and Statistics)

Abstract

:
Let F ( ν j ) = { Q m j ν j , m j ( m ν j , m + ν j ) } , j = 1 , 2 , be two Cauchy–Stieltjes Kernel (CSK) families induced by non-degenerate compactly supported probability measures ν 1 and ν 2 . Introduce the set of measures F = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) a n d m 2 ( m ν 2 , m + ν 2 ) } . We show that if F remains a CSK family, (i.e., F = F ( μ ) where μ is a non-degenerate compactly supported measure), then the measures μ , ν 1 and ν 2 are of the Marchenko–Pastur type measure up to affinity. A similar conclusion is obtained if we substitute (in the definition of F ) the additive free convolution ⊞ by the additive Boolean convolution ⊎. The cases where the additive free convolution ⊞ is replaced (in the definition of F ) by the multiplicative free convolution ⊠ or the multiplicative Boolean convolution ⨃ are also studied.

1. Introduction

Convolution stability is an important concept in probability theory, particularly in the study of probability laws and stochastic processes. It represents a fundamental concept that helps understand how certain probability laws behave under addition and has profound implications in various practical applications.
In [1], the authors studied the stability (with respect to classical convolution) of natural exponential families (NEFs) of probability measures: If the convolution of NEFs on R d remains a NEF, then the NEFs are all Poisson–Gaussian, up to affinity. This result is a generalization of the one-dimensional versions demonstrated in [2] in the case of two NEFs, and in [3] for more than two NEFs.
In non-commutative probability, the concept of Cauchy–Stieltjes kernel (CSK) families is newly defined. It concerns families of probability measures introduced in a manner similar to that of NEFs by involving the Cauchy–Stieltjes kernel (CSK) 1 1 ϑ y instead of the exponential kernel exp ( ϑ y ) . The CSK families have been studied in [4] for measures with compact support. Further properties are proven in [5], by exploring measures with a one-sided support boundary, such as the one mentioned above. In this article, we study the stability of CSK families with respect to free and Boolean convolutions product. This stability property subsumes the CSK family induced by the Marchenko–Pastur type law and provides some new variance functions with non-usual expressions. The approach used in this article extends the understanding of CSK families and their stability in the non-commutative probability framework. In the rest of this section, after giving some basic concepts about CSK families, the purpose of this article is presented in more detail.
Let μ be a (non-degenerate) probability measure with a one-sided support boundary from above. Then,
M μ ( ϑ ) = 1 1 ϑ y μ ( d y )
converges ∀ ϑ [ 0 , ϑ + μ ) with 1 / ϑ + μ = max { sup supp ( μ ) , 0 } . The CSK family induced by μ is the family of probabilities
F + ( μ ) = { P ( ϑ , μ ) ( d y ) = 1 M μ ( ϑ ) ( 1 ϑ y ) μ ( d y ) : ϑ ( 0 , ϑ + μ ) } .
The map ϑ K μ ( ϑ ) = y P ( ϑ , μ ) ( d y ) represent mean function. It is increasing strictly on ( 0 , ϑ + μ ) , see [5]. The interval ( m 0 μ , m + μ ) = K μ ( ( 0 , ϑ + μ ) ) is said to be the mean domain of F + ( μ ) . This leads to the mean re-parametrization of F + ( μ ) . Let ψ μ ( · ) the inverse function of K μ ( · ) and for x ( m 0 μ , m + μ ) , write Q x μ ( d y ) = P ( ψ μ ( x ) , μ ) ( d y ) . We obtain
F + ( μ ) = { Q x μ ( d y ) : x ( m 0 μ , m + μ ) } .
Consider
B = B ( μ ) = max { 0 , sup supp ( μ ) } = 1 / ϑ + [ 0 , ) .
It is known from [5] that
m 0 μ = lim ϑ 0 + K μ ( ϑ ) and m + μ = B lim w B + 1 G μ ( w ) ,
where
G μ ( w ) = 1 w y μ ( d y ) , w C supp ( μ ) .
is the Cauchy transform of μ .
The CSK family induced by a measure μ with a support bounded from below can also be introduced. It is denoted F ( μ ) and parameterized by ϑ ( ϑ μ , 0 ) , where ϑ μ is either or 1 / A ( μ ) with A = A ( μ ) = min { inf s u p p ( μ ) , 0 } . The mean domain for F ( μ ) is ( m μ , m 0 μ ) with m μ = A 1 / G μ ( A ) . If μ is compactly supported, then F ( μ ) = F + ( μ ) F ( μ ) { μ } is the two-sided CSK family.
The variance function (VF) is defined by (see [4])
x V μ ( x ) = ( ζ x ) 2 Q x μ ( d ζ ) .
If the first moment of μ does not exist, then all the laws in F + ( μ ) have infinite variance. The notion of pseudo-variance function V μ ( · ) is defined in [5] as
V μ ( x ) = x 1 ψ μ ( x ) x ,
If m 0 μ = y μ ( d y ) exists finitely, then the VF exists (see [5]) and
V μ ( x ) = x x m 0 μ V μ ( x ) .
Remark 1.
(i) V μ ( · ) determine μ: Consider ϖ = ϖ ( x ) = x + V μ ( x ) x , then
G μ ( ϖ ) = x V μ ( x ) .
Furthermore, if m 0 μ is finite, then
G μ ( ϖ ) = x m 0 μ V μ ( x ) .
Consequently, V μ ( · ) and m 0 μ determine μ.
(ii) 
Let χ ( μ ) be the image of μ by χ : y ξ y + ι where ξ 0 and ι R . Then,m close to m 0 χ ( μ ) = χ ( m 0 μ ) = ξ m 0 μ + ι ,
V χ ( μ ) ( x ) = ξ 2 x x ι V μ x ι ξ .
If the VF exists, then
V χ ( μ ) ( x ) = ξ 2 V μ x ι ξ .
(iii) 
For a 0 , the Marchenko–Pastur law is
MP ( a ) ( d s ) = 4 ( s a ) 2 2 π ( 1 + a s ) 1 ( a 2 , a + 2 ) ( s ) d s + ( 1 1 / a 2 ) δ 1 / a ( d s ) ,
with m 0 MP ( a ) = 0 . We know that
V MP ( a ) ( m ) = 1 + a m , m ( m MP ( a ) , m + MP ( a ) ) = ( 1 , 1 ) .
See ([4], [Theorem 3.2]), for more details.
Now, we recall the definition of additive free convolution (see [6] for more details) and we present the purpose of this article. Denote by P (respectively, by P + , P c , P c + ) the set of non-degenerate probability measures on R (respectively, the subset of measures from P concentrated on the positive real line, with compact support, concentrated on the positive real line and compactly supported). The additive free convolution of μ and ν P is the measure denoted as μ ν , satisfying
R μ ν ( ξ ) = R μ ( ξ ) + R ν ( ξ ) ,
where the free cumulant transformation, R μ , of μ is
R μ ( G μ ( ξ ) ) = ξ 1 G μ ( ξ ) , ξ clos   to   0 .
The objective in this article related to the ⊞-convolution is as follows: Let F ( ν j ) = { Q m j ν j , m j ( m ν j , m + ν j ) } , j = 1 , 2 , be two CSK families induced by ν 1 and ν 2 P c . Introduce the family of probabilities
F = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) and m 2 ( m ν 2 , m + ν 2 ) } .
We demonstrate that if F remains a CSK family, (i.e., F = F ( μ ) for μ P c ), then the measures μ , ν 1 and ν 2 are of the Marchenko–Pastur type law up to affinity. We obtain the same conclusion if we replace (in the definition of F ) the additive free convolution ⊞ by the additive Boolean convolution ⊎. The cases where the additive free convolution ⊞ is replaced (in the definition of F ) by the multiplicative free convolution ⊠ or the multiplicative Boolean convolution ⨃ are also studied. This provides some new VFs with non-usual forms.

2. Main Results

2.1. Stability of CSK Families by Free Additive Convolution Product

Theorem 1.
Let ν 1 , ν 2 P c . Introduce the family of measures
F = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) a n d m 2 ( m ν 2 , m + ν 2 ) } .
If F remains a CSK family, (i.e., F = F ( μ ) for μ P c ), then μ, ν 1 and ν 2 are of the Marchenko–Pastur type law up to affinity.
Proof. 
We omit the proof because it is the same when considering t = 1 in the proof of ([7], [Theorem 2]). □

2.2. Stability of CSK Families by Boolean Additive Convolution Product

For μ , ν P , the additive Boolean convolution μ ν is the probability measure introduced by
E μ ν ( ξ ) = E μ ( ξ ) + E ν ( ξ ) , for ξ C + ,
where
E μ ( ξ ) = ξ 1 G μ ( ξ ) , for ξ C + .
See [8] for more details.
Theorem 2.
Let ν 1 , ν 2 P c . Introduce the family of probabilities
H = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) a n d m 2 ( m ν 2 , m + ν 2 ) } .
If H remains a CSK family, (i.e., H = F ( σ ) for σ P c ), then σ, ν 1 and ν 2 are of the Marchenko–Pastur type law up to affinity.
Proof. 
Assume that H = F ( σ ) for σ P c , (we may suppose, without loss of generality, that 0 ( m σ , m + σ ) ). Then, ∀ m 1 ( m ν 1 , m + ν 1 ) and ∀ m 2 ( m ν 2 , m + ν 2 ) , there is t ( m σ , m + σ ) , so that
Q t σ = Q m 1 ν 1 Q m 2 ν 2 .
Or equivalently, ∀ ξ large enough
E Q t σ ( ξ ) = E Q m 1 ν 1 ( ξ ) + E Q m 2 ν 2 ( ξ ) .
For σ P c , the Boolean cumulant transform of Q t σ P c may be expressed as
E Q t σ ( ξ ) = r 1 ( Q t σ ) + r 2 ( Q t σ ) 1 ξ + 1 ξ ε 1 ξ with ε 1 ξ ξ + 0 .
where r 1 ( Q t σ ) and r 2 ( Q t σ ) denotes, respectively, the first and the second Boolean cumulants of Q t σ . It follows from ([9], [Equation 3.2]) that
r 1 ( Q t σ ) = t and r 2 ( Q t σ ) = variance   of Q t σ = V σ ( t ) .
So, relation (2) reduces to
E Q t σ ( ξ ) = t + V σ ( t ) 1 ξ + 1 ξ ε 1 ξ with ε 1 ξ ξ + 0 .
Similarly, the Boolean cumulant transforms of Q m 1 ν 1 and Q m 2 ν 2 P c may be written as
E Q m 1 ν 1 ( ξ ) = m 1 + V ν 1 ( m 1 ) 1 ξ + 1 ξ ε 1 1 ξ with ε 1 1 ξ ξ + 0
and
E Q m 2 ν 2 ( ξ ) = m 2 + V ν 2 ( m 2 ) 1 ξ + 1 ξ ε 2 1 ξ with ε 2 1 ξ ξ + 0 .
Equation (1), together with (3), (4), (5) implies that
t + V σ ( t ) 1 ξ + 1 ξ ε 1 ξ = m 1 + m 2 + ( V ν 1 ( m 1 ) + V ν 2 ( m 2 ) ) 1 ξ + 1 ξ ε 1 1 ξ + ε 2 1 ξ .
When ξ + in (6), we obtain t = m 1 + m 2 . So, Equation (6) is reduced to
V σ ( m 1 + m 2 ) 1 ξ + 1 ξ ε 1 ξ = ( V ν 1 ( m 1 ) + V ν 2 ( m 2 ) ) 1 ξ + 1 ξ ε 1 1 ξ + ε 2 1 ξ .
Multiplying by ξ in both sides of (7) and letting ξ + , we obtain
V σ ( m 1 + m 2 ) = V ν 1 ( m 1 ) + V ν 2 ( m 2 ) .
According to [10], we obtain
V σ ( m ) = V ν 1 ( m ) = γ m + ς and V ν 2 ( m ) = γ m for γ 0 and ς > 0 .
The parameter ς = V σ ( 0 ) should be strictly positive. In addition, m 0 ν 2 0 . Otherwise m V ν 2 ( m ) = λ m cannot be a VF, see ([11], [page 6]). Clearly from (8), measures σ , ν 1 and ν 2 are of Marchenko–Pastur type law up to affinity. □

2.3. Stability of CSK Families by Multiplicative Free Convolution Product

For μ P + , the S -transform is defined by
M μ z 1 + z S μ ( z ) = 1 + z , z close   to   0 .
The product of S -transforms is an S -transform and defines, for ν 1 , ν 2 P + , the multiplicative free convolution ν 1 ν 2 , by S ν 1 ν 2 ( z ) = S ν 1 ( z ) S ν 2 ( z ) . See [6,12] for more details.
Theorem 3.
Let ν 1 , ν 2 P c + . Introduce the family of probabilities
Z = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) a n d m 2 ( m ν 2 , m + ν 2 ) } .
If Z remains a CSK family, (i.e., Z = F ( σ ) for σ P c + ), then σ, ν 1 and ν 2 are such that
V ν 1 ( m ) = m 2 ( c ln ( m ) + a 1 ) , m ( m ν 1 , m + ν 2 ) ,
V ν 2 ( m ) = m 2 ( c ln ( m ) + a 2 ) , m ( m ν 2 , m + ν 2 ) ,
and
V σ ( m ) = m 2 ( c ln ( m ) + a 1 + a 2 ) , m ( m σ , m + σ ) .
a 1 and a 2 are positive strictly and c ( κ , + ) with κ = max { a 1 ln ( m + ν 1 ) , a 2 ln ( m + ν 2 ) , ( a 1 + a 2 ) ln ( m + σ ) } .
Proof. 
Suppose that Z = F ( σ ) for some σ P c + . Then ∀ m 1 ( m ν 1 , m + ν 1 ) and ∀ m 2 ( m ν 2 , m + ν 2 ) , there is r ( m σ , m + σ ) , so that
Q r σ = Q m 1 ν 1 Q m 2 ν 2 .
That is, ∀z close to 0
S Q r σ ( z ) = S Q m 1 ν 1 ( z ) S Q m 2 ν 2 ( z ) .
We know that lim z 0 S σ ( z ) = 1 m 0 σ . Thus
1 r = 1 m 0 Q r σ = lim z 0 S Q r σ ( z ) = lim z 0 S Q m 1 ν 1 ( z ) S Q m 2 ν 2 ( z ) = 1 m 1 m 2 .
Then, r = m 1 m 2 . Using ([13], [Theorem 3.3]), the S -transforms of measures Q m 1 m 2 σ , Q m 1 ν 1 and Q m 2 ν 2 may be written as
S Q m 1 m 2 σ ( z ) = 1 m 1 m 2 V σ ( m 1 m 2 ) ( m 1 m 2 ) 3 z + + z ϵ ( z ) , where ϵ ( z ) z 0 0
S Q m 1 ν 1 ( z ) = 1 m 1 V ν 1 ( m 1 ) m 1 3 z + z ϵ 1 ( z ) , where ϵ 1 ( z ) z 0 0
and
S Q m 2 ν 2 ( z ) = 1 m 2 V ν 2 ( m 2 ) m 2 3 z + z ϵ 2 ( z ) , where ϵ 2 ( z ) z 0 0 .
Combining (12), (13), (14) and (15), we obtain
V σ ( m 1 m 2 ) ( m 1 m 2 ) 2 = V ν 1 ( m 1 ) m 1 2 + V ν 2 ( m 2 ) m 2 2 .
On the other hand, it is well known that the solutions the functional equation
f ( x 1 x 2 ) = f 1 ( x 1 ) + f 2 ( x 2 )
where the functions f ( · ) , f 1 ( · ) and f 2 ( · ) are continuous and x 1 , x 2 0 , is
f ( x ) = c ln ( x ) + a 1 + a 2 and f i ( x ) = c ln ( x ) + a i , i = 1 , 2 .
c, a 1 and a 2 arbitrary constants.
Thus, Equation (16) implies that the VFs of F ( ν 1 ) , F ( ν 2 ) and F ( σ ) are given, respectively, by (9), (10) and (11). □
Remark 2.
(i) Up to affinity, we may suppose that ( m ν i , m + ν i ) ( 1 , + ) , i = 1, 2.
(ii) 
For c = 0 , we have V ν 1 ( m ) = a 1 m 2 , V ν 2 ( m ) = a 2 m 2 and V σ ( m ) = ( a 1 + a 2 ) m 2 . Thus, the parameters a 1 and a 2 must strictly be positive. In that case, the measures ν 1 , ν 2 and σ are of free Gamma type law up to affinity. For more details about free Gamma CSK family, see ([4], [Theorem 3.2]).
(iii) 
We justify the choice of the parameter c, to guarantee the positivity of the variance functions given (9), (10) and (11). That is c ( κ , + ) , where κ = max { a 1 ln ( m + ν 1 ) , a 2 ln ( m + ν 2 ) , ( a 1 + a 2 ) ln ( m + σ ) } .

2.4. Stability of CSK Families by Boolean Multiplicative Convolution Product

In [14], a multiplicative concept of Boolean additive convolution was introduced. For λ P + , the map
Φ λ ( ξ ) = 0 + x ξ 1 x ξ λ ( d x ) , ξ C R +
is univalent in i C + . It is known that the circle with diameter ( λ ( { 0 } ) 1 , 0 ) contains Φ λ ( i C + ) . In addition, R Φ λ ( i C + ) = ( λ ( { 0 } ) 1 , 0 ) . We also have Φ λ ( ξ 1 ) = ξ G λ ( ξ ) 1 . Following [15], the η -transformation of λ P + is provided as follows:
η λ : C R + C R + ; ξ η λ ( ξ ) = Φ λ ( ξ ) 1 + Φ λ ( ξ ) .
As is well known, η λ ( ( , 0 ) ) ( , 0 ) , lim ξ 0 , ξ < 0 η λ ( ξ ) = η λ ( 0 ) = 0 and η λ ( ξ ¯ ) = η λ ( ξ ) ¯ , for ξ C R + . Furthermore, arg ( ξ ) arg ( η λ ( ξ ) ) < π , for ξ C + .
The transform
B λ ( ξ ) = ξ η λ ( ξ )
is well defined for ξ C R + . The Boolean multiplicative convolution of τ and λ P + is the unique probability measure in P + , represented as λ τ , that fulfills
B λ τ ( ξ ) = B λ ( ξ ) B τ ( ξ ) , f o r ξ C R + .
For τ , λ P + satisfying
(i)
arg ( η λ ( ξ ) ) + arg ( η τ ( ξ ) ) arg ( ξ ) < π f o r ξ C + ( , 0 )
(ii)
at least one of the first moments of one of the probabilities λ or τ is finite, the measure λ τ P + is well defined.
Theorem 4.
Let ν 1 , ν 2 P c + . Introduce the family of probabilities
L = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) a n d m 2 ( m ν 2 , m + ν 2 ) } .
If L remains a CSK family, (i.e., L = F ( σ ) for σ P c + ), then ν 1 , ν 2 and σ are such that
V ν 1 ( m ) = m ( d ln ( m ) + b 1 ) , m ( m ν 1 , m + ν 1 ) ,
V ν 2 ( m ) = m ( d ln ( m ) + b 2 ) , m ( m ν 2 , m + ν 2 ) ,
and
V σ ( m ) = m ( d ln ( m ) + b 1 + b 2 ) , m ( m σ , m + σ ) .
b 1 and b 2 are strictly positive and d ( υ , + ) with υ = max { b 1 ln ( m + ν 1 ) , b 2 ln ( m + ν 2 ) , ( b 1 + b 2 ) ln ( m + σ ) } .
Proof. 
Suppose that L = F ( σ ) for some σ P c + . Then, ∀ m 1 ( m ν 1 , m + ν 1 ) and ∀ m 2 ( m ν 2 , m + ν 2 ) , there is r ( m σ , m + σ ) , so that
Q r σ = Q m 1 ν 1 Q m 2 ν 2 .
That is, ∀z close to 0
B Q r σ ( z ) = B Q m 1 ν 1 ( z ) B Q m 2 ν 2 ( z ) .
On the other hand, the multiplicative Boolean cumulant transform is related to the additive Boolean cumulant transform as
1 B Q r σ ( z ) = E Q r σ ( 1 / z ) .
Based on (23), (3), (4) and (5), one can obtain
1 B Q r σ ( z ) = r + V σ ( r ) z + z ε ( z ) with ε ( z ) z 0 0 ,
1 B Q m 1 ν 1 ( z ) = m 1 + V ν 1 ( m 1 ) z + z ε 1 ( z ) with ε 1 ( z ) z 0 0
and
1 B Q m 2 ν 2 ( z ) = m 2 + V ν 2 ( m 2 ) z + z ε 2 ( z ) with ε 2 ( z ) z 0 0 .
Combining (22) with (24), (25) and (26), we obtain
r + V σ ( r ) z + z ε ( z ) = m 1 m 2 + ( m 1 V ν 2 ( m 2 ) + m 2 V ν 1 ( m 1 ) ) z + z ε 3 ( z ) with ε 3 ( z ) z 0 0 .
When z 0 in (27), we obtain r = m 1 m 2 . Thus, relation (27) reduces to
V σ ( m 1 m 2 ) z + z ε ( z ) = ( m 1 V ν 2 ( m 2 ) + m 2 V ν 1 ( m 1 ) ) z + z ε 3 ( z ) with ε 3 ( z ) z 0 0 .
Dividing by z in both sides of (28) and let z 0 , one obtains V σ ( m 1 m 2 ) m 1 m 2 = V ν 1 ( m 1 ) m 1 + V ν 2 ( m 2 ) m 2 . Recalling (17), we obtain from (18)
V ν 1 ( m ) m = d ln ( m ) + b 1 ,
V ν 2 ( m ) m = d ln ( m ) + b 2
and
V σ ( m ) m = d ln ( m ) + b 1 + b 2 .
Thus, measures ν 1 , ν 2 and σ are characterized by their corresponding VFs provided, respectively, by (19), (20) and (21). □
Remark 3.
(i) Up to affinity, we may suppose that ( m ν i , m + ν i ) ( 1 , + ) , i = 1, 2.
(ii) 
Note that for d = 0 , we have V ν 1 ( m ) = b 1 m , V ν 2 ( m ) = b 2 m and V σ ( m ) = ( b 1 + b 2 ) m . Thus, the parameters b 1 and b 2 must be positive strictly. In that case the measures ν 1 , ν 2 and σ are of Marchenko–Pastur type law up to affinity.
(iii) 
We justify the choice of the parameter υ, to guarantee the positivity of the variance functions given in (19), (20) and (21). That is d ( υ , + ) where υ = max { b 1 ln ( m + ν 1 ) , b 2 ln ( m + ν 2 ) , ( b 1 + b 2 ) ln ( m + σ ) } .

3. Conclusions

This article provides significant insights into the stability of CSK families under both free and Boolean convolutions. Let F ( ν j ) = { Q m j ν j , m j ( m ν j , m + ν j ) } , j = 1 , 2 , be two CSK families induced by ν 1 and ν 2 P c . Introduce the family of measures
H = F ( ν 1 ) F ( ν 2 ) = { Q m 1 ν 1 Q m 2 ν 2 , m 1 ( m ν 1 , m + ν 1 ) and m 2 ( m ν 2 , m + ν 2 ) } .
We have proved that if H remains a CSK family, (i.e., H = F ( μ ) for μ P c ), then μ , ν 1 and ν 2 are of the Marchenko–Pastur type law up to affinity. We obtain the same conclusion if we replace (in the definition of H ) the additive Boolean convolution ⊎ with the additive free convolution ⊞. Other cases are also studied where the additive Boolean convolution is replaced (in the definition of H ) by the multiplicative free convolution ⊠ or the multiplicative Boolean convolution ⨃. This provides to some new VFs with logarithmic terms. In fact, in many cases, the variance of the response variable increases with the mean. A logarithmic transformation can help stabilize the variance, making it more constant across different levels of the predictor variables. Furthermore, the logarithmic function has desirable mathematical properties that can simplify the analysis, such as transforming multiplicative relationships into additive ones, which are easier to handle in regression analysis. In addition, in some cases, there might be theoretical reasons based on the underlying processes generating the data that suggest a logarithmic relationship between the mean and variance.

Author Contributions

Conceptualization, A.R.A.A. and S.S.A.; methodology, R.F.; validation, A.R.A.A. and S.S.A.; investigation, S.S.A.; resources, S.S.A. and R.F.; writing—original draft, R.F.; writing—review and editing, R.F.; visualization, A.R.A.A. and S.S.A.; supervision, R.F.; project administration, A.R.A.A.; funding acquisition, A.R.A.A. All authors have read and agreed to the published version of the manuscript.

Funding

This research is funded by the Deanship of Graduate Studies and Scientific Research at Jouf University through the Fast-Track Research Funding Program.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Koudou, A.E.; Pommeret, D. A Characterization of Poisson-Gaussian Families by Convolution-Stability. J. Multivar. Anal. 2002, 81, 120–127. [Google Scholar] [CrossRef]
  2. Letac, G. Lectures on Natural Exponential Families and Their Variance Functions; Instituto de Matemtica Pura e Aplicada: Monografias de Matemtica, Rio de Janeiro, Brazil, 1992; Volume 50. [Google Scholar]
  3. Pommeret, D. Stabilité des familles exponentielles naturelles par convolution. C. R. Acad. Sci. Sr. I 1999, 328, 929–933. [Google Scholar]
  4. Bryc, W. Free exponential families as kernel families. Demonstr. Math. 2009, 42, 657–672. [Google Scholar] [CrossRef]
  5. Bryc, W.; Hassairi, A. One-sided Cauchy-Stieltjes kernel families. J. Theoret. Probab. 2011, 24, 577–594. [Google Scholar] [CrossRef]
  6. Bercovici, H.; Voiculescu, D. Free convolution of measures with unbounded support. Indiana Univ. Math. J. 1993, 42, 733–773. [Google Scholar] [CrossRef]
  7. Alanzi, A.R.A.; Alqasem, O.A.; Elwahab, M.E.A.; Fakhfakh, R. Some Results on the Free Poisson Distribution. Axioms 2024, 13, 496. [Google Scholar] [CrossRef]
  8. Speicher, R.; Woroudi, R. Boolean convolution. Fields Inst. Commun. 1997, 12, 26727. [Google Scholar] [CrossRef]
  9. Fakhfakh, R. Variance function of boolean additive convolution. Stat. Probab. Lett. 2020, 163, 108777. [Google Scholar] [CrossRef]
  10. Friedman, D. The functional equation f(x+y)=g(x)+h(y). Am. Math. Mon. 1962, 69, 769–772. [Google Scholar] [CrossRef]
  11. Fakhfakh, R. A characterization of the Marchenko-Pastur probability measure. Stat. Probab. Lett. 2022, 19, 109660. [Google Scholar] [CrossRef]
  12. Belinschi, S.T. Complex Analysis Methods in Noncommutative Probability; ProQuest LLC.: Ann Arbor, MI, USA, 2005. [Google Scholar]
  13. Noriyoshi, S.; Hiroaki, Y. New limit theorems related to free multiplicative convolution. Stud. Math. 2013, 214, 251–264. [Google Scholar]
  14. Bercovici, H. On Boolean convolutions. In Operator Theory 20; MR2276927l; Theta: Bucharest, Romania, 2006; Volume 6, pp. 7–13. [Google Scholar]
  15. Anshelevich, M.; Wang, J.C.; Zhong, P. Local limit theorems for multiplicative free convolutions. J. Funct. Anal. 2014, 267, 3469–3499. [Google Scholar] [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Alanzi, A.R.A.; Alshqaq, S.S.; Fakhfakh, R. Stability of Cauchy–Stieltjes Kernel Families by Free and Boolean Convolutions Product. Mathematics 2024, 12, 3465. https://doi.org/10.3390/math12223465

AMA Style

Alanzi ARA, Alshqaq SS, Fakhfakh R. Stability of Cauchy–Stieltjes Kernel Families by Free and Boolean Convolutions Product. Mathematics. 2024; 12(22):3465. https://doi.org/10.3390/math12223465

Chicago/Turabian Style

Alanzi, Ayed. R. A., Shokrya S. Alshqaq, and Raouf Fakhfakh. 2024. "Stability of Cauchy–Stieltjes Kernel Families by Free and Boolean Convolutions Product" Mathematics 12, no. 22: 3465. https://doi.org/10.3390/math12223465

APA Style

Alanzi, A. R. A., Alshqaq, S. S., & Fakhfakh, R. (2024). Stability of Cauchy–Stieltjes Kernel Families by Free and Boolean Convolutions Product. Mathematics, 12(22), 3465. https://doi.org/10.3390/math12223465

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop