
Count data modeling and its practical applications have garnered significant attention in recent research, owing to its relevance in a wide range of fields. This study specifically explores a novel discrete distribution characterized by two parameters, which is derived using the survival discretization method. The statistical properties of this distribution are thoroughly explained in closed forms, with several key mathematical attributes also derived. These characteristics underscore the distribution's effectiveness in modeling data that exhibit (right-skewed) asymmetry and have extended heavy tails, making it particularly suitable for such real-world applications. Furthermore, the failure rate function corresponding to this distribution is particularly appropriate for scenarios characterized by an increasing or bathtub-shaped failure rate over time. The model is also highly versatile, offering valuable insights into probabilistic modeling for datasets that display over dispersion, under dispersion, or equi dispersion. The study introduces several estimation techniques, including the maximum product of spacings, Anderson–Darling, right–tail Anderson–Darling, maximum likelihood estimation, least squares, weighted least squares, Cramer–Von–Mises, and percentile methods. Each of these methods is explained in detail, providing a comprehensive understanding of their application. A ranking simulation study is conducted to evaluate the performance of these estimators across varying sample sizes, using ranking techniques to identify the most effective estimator in different scenarios. The analysis of real-world datasets from biotechnology and industrial engineering further demonstrates the practical utility and relevance of the proposed model. The results highlight the model's ability to offer accurate and insightful analyses, reinforcing its significance in count data modeling and its wide-ranging applications.
Citation: Mohamed S. Algolam, Mohamed S. Eliwa, Mohamed El-Dawoody, Mahmoud El-Morshedy. A discrete extension of the Xgamma random variable: mathematical framework, estimation methods, simulation ranking, and applications to radiation biology and industrial engineering data[J]. AIMS Mathematics, 2025, 10(3): 6069-6101. doi: 10.3934/math.2025277
[1] | A. M. Elaiw, A. S. Shflot, A. D. Hobiny . Stability analysis of SARS-CoV-2/HTLV-I coinfection dynamics model. AIMS Mathematics, 2023, 8(3): 6136-6166. doi: 10.3934/math.2023310 |
[2] | Ruiqing Shi, Yihong Zhang . Dynamic analysis and optimal control of a fractional order HIV/HTLV co-infection model with HIV-specific antibody immune response. AIMS Mathematics, 2024, 9(4): 9455-9493. doi: 10.3934/math.2024462 |
[3] | E. A. Almohaimeed, A. M. Elaiw, A. D. Hobiny . Modeling HTLV-1 and HTLV-2 co-infection dynamics. AIMS Mathematics, 2025, 10(3): 5696-5730. doi: 10.3934/math.2025263 |
[4] | A. M. Elaiw, E. A. Almohaimeed, A. D. Hobiny . Stability of HHV-8 and HIV-1 co-infection model with latent reservoirs and multiple distributed delays. AIMS Mathematics, 2024, 9(7): 19195-19239. doi: 10.3934/math.2024936 |
[5] | Ru Meng, Yantao Luo, Tingting Zheng . Stability analysis for a HIV model with cell-to-cell transmission, two immune responses and induced apoptosis. AIMS Mathematics, 2024, 9(6): 14786-14806. doi: 10.3934/math.2024719 |
[6] | Mohammed H. Alharbi . HIV dynamics in a periodic environment with general transmission rates. AIMS Mathematics, 2024, 9(11): 31393-31413. doi: 10.3934/math.20241512 |
[7] | Hanan S. Gafel, Saima Rashid, Sayed K. Elagan . Novel codynamics of the HIV-1/HTLV-Ⅰ model involving humoral immune response and cellular outbreak: A new approach to probability density functions and fractional operators. AIMS Mathematics, 2023, 8(12): 28246-28279. doi: 10.3934/math.20231446 |
[8] | Afnan Al Agha, Hakim Al Garalleh . Oncolysis by SARS-CoV-2: modeling and analysis. AIMS Mathematics, 2024, 9(3): 7212-7252. doi: 10.3934/math.2024351 |
[9] | Liang Hong, Jie Li, Libin Rong, Xia Wang . Global dynamics of a delayed model with cytokine-enhanced viral infection and cell-to-cell transmission. AIMS Mathematics, 2024, 9(6): 16280-16296. doi: 10.3934/math.2024788 |
[10] | Attaullah, Sultan Alyobi, Mansour F. Yassen . A study on the transmission and dynamical behavior of an HIV/AIDS epidemic model with a cure rate. AIMS Mathematics, 2022, 7(9): 17507-17528. doi: 10.3934/math.2022965 |
Count data modeling and its practical applications have garnered significant attention in recent research, owing to its relevance in a wide range of fields. This study specifically explores a novel discrete distribution characterized by two parameters, which is derived using the survival discretization method. The statistical properties of this distribution are thoroughly explained in closed forms, with several key mathematical attributes also derived. These characteristics underscore the distribution's effectiveness in modeling data that exhibit (right-skewed) asymmetry and have extended heavy tails, making it particularly suitable for such real-world applications. Furthermore, the failure rate function corresponding to this distribution is particularly appropriate for scenarios characterized by an increasing or bathtub-shaped failure rate over time. The model is also highly versatile, offering valuable insights into probabilistic modeling for datasets that display over dispersion, under dispersion, or equi dispersion. The study introduces several estimation techniques, including the maximum product of spacings, Anderson–Darling, right–tail Anderson–Darling, maximum likelihood estimation, least squares, weighted least squares, Cramer–Von–Mises, and percentile methods. Each of these methods is explained in detail, providing a comprehensive understanding of their application. A ranking simulation study is conducted to evaluate the performance of these estimators across varying sample sizes, using ranking techniques to identify the most effective estimator in different scenarios. The analysis of real-world datasets from biotechnology and industrial engineering further demonstrates the practical utility and relevance of the proposed model. The results highlight the model's ability to offer accurate and insightful analyses, reinforcing its significance in count data modeling and its wide-ranging applications.
Overfishing occurs when fish are harvested at a rate faster than the resource can replenish itself. This practice has consequences that extend beyond the depletion of single-species fish populations, potentially leading to global food insecurity, ecological extinction, and other devastating effects[1,2,3,4]. Marine ecosystems and their associated livelihoods worldwide are threatened by overfishing. Nicholas et al. [5] have shown that overfishing is pushing over one-third of all sharks and rays towards global extinction. They suggest that marine life could face a future similar to that of land biodiversity, where human pressures have resulted in the loss of numerous species, possibly triggering a sixth mass extinction. According to IUCN Red List criteria [6], they estimate that one-quarter of 1,041 chondrichthyan fishes are threatened due to overfishing (both targeted and incidental). Therefore, it is urgent to establish the optimal harvesting strategy (OHS) for overfishing to prevent extinctions and promote population recovery. Consequently, investigating OHS has significant implications for marine ecosystems.
Mathematical models are widely utilized to analyze the OHS of populations [7,8,9,10,11,12,13]. Clark [7] proposes several deterministic population systems with harvesting and establishes the OHS. Martin and Ruan [8] investigate the predator-prey models with delay and prey harvesting. However, nature is full of random phenomena, and various populations in the ecosystem experience different forms of random perturbations. May [9] indicates that environmental noise affects competition coefficient, environmental capacity, the growth rate and other parameters of population systems. Lande et al.[10] demonstrate that the extinction of numerous species caused by over-harvesting may be largely due to the ignorance of randomness. Therefore, it is crucial to investigate the OHS of population system in random environment. Li and Wang [11] establish a general stochastic logistic population model with white noises and explore the optimal harvesting policy. Wang et al.[12] propose a stochastic predator-prey model with time delays and shows how randomness influences the extinction and OHS. Therefore, exploring the OHS of stochastic marine population systems is meaningful.
Lévy noise, another type of environmental perturbations, is considered below. During the process of species evolution, population growth is often affected by random disturbances in the environment, such as toxic pollutants, epidemics, earthquakes, floods and other factors [14,15,16,17]. For example, the Sandoz chemical accident in 1986 resulted in the death of over 600,000 fish in the Rhine [14], while the 2010 Gulf of Mexico oil spill caused a devastating impact on marine life in the Gulf of Mexico [15]. Bao et al. [16] propose that non-Gaussian Lévy noise is suitable for concluding these phenomena and revealed that Lévy noise can inhibit the explosive increase of population. Zhang et al. [17] also proposed to use Lévy noise to describe the random perturbations.
A lot of work has been done to study the stochastic food chain population dynamics model so far. Our previous work has investigated the dynamics of a stochastic three species food chain model in [18], the dynamics of a four-species stochastic food chain model with delays in [19] and the dynamic behaviors of a stochastic n-species food chain model with Lévy noises and white noises in [20] as follows:
{dx1(t)=x1(t)[r1−a11x1(t)−a12x2(t)]dt+σ1x1(t)dB1(t)+∫Zx1(t−)γ1(u)˜N(dt,du),dxi(t)=xi(t)[ri−aiixi(t)+aii−1xi−1(t)−aii+1xi+1(t)]dt+σixi(t)dBi(t)+∫Zxi(t−)γi(u)˜N(dt,du),2≤i≤n−1,dxn(t)=xn(t)[rn−annxn(t)+ann−1xn−1(t)]dt+σnxn(t)dBn(t)+∫Zxn(t−)γn(u)˜N(dt,du). | (1.1) |
By introducing a new research method, that is algebraic approach, the sufficient and necessary criteria for the stochastic extinction, persistence in mean and global stability are established in [20]. These conclusions directly generalize some research results on three and four species stochastic food chain dynamics models in [18,19]. However, due to the high dimensionality of model (1.1) and technical difficulties, the stability in distribution and optimal harvesting problems have not been discussed until now. In this paper, we consider the effects of harvesting and Lévy noises, then address these issues by proposing a stochastic n-species food chain model as follows:
{dx1(t)=x1(t)[r1−h1−a11x1(t)−a12x2(t)]dt+σ1x1(t)dB1(t)+∫Zx1(t−)γ1(u)˜N(dt,du),dxi(t)=xi(t)[ri−hi−aiixi(t)+aii−1xi−1(t)−aii+1xi+1(t)]dt+σixi(t)dBi(t)+∫Zxi(t−)γi(u)˜N(dt,du),2≤i≤n−1,dxn(t)=xn(t)[rn−hi−annxn(t)+ann−1xn−1(t)]dt+σnxn(t)dBn(t)+∫Zxn(t−)γn(u)˜N(dt,du). | (1.2) |
On the basis of the research results in [20], our primary objective is to establish criteria for the stability in distribution of model (1.2). Additionally, we aim to establish criteria for the optimal harvesting strategy in the sense of sustainable yield. Lastly, we will verify the theoretical results through numerical examples and propose some open questions and propositions.
This paper is organized as follows. In Section 2, we provide the basic assumption, and further introduce several important lemmas that will be used to prove the major results. In Section 3, we state and prove the stability in distribution of model (1.2). In Section 4, we outline the non-existence and existence criteria for OHS and MESY. Finally, in Section 5, we illustrate our major results through numerical examples.
Denote by (Ω,{Ft}t≥0,P) the complete probability space, and its filtration {Ft}t≥0 satisfies the usual conditions. In this paper, we assume that model (1.2) is defined on the probability space (Ω,{Ft}t≥0,P). In addition, for an integrable function f(t) defined for t∈[0,+∞), we define the mean value of f(t) on interval [0,t] for any t>0 by ⟨f(t)⟩=1t∫t0f(s)ds.
In models (1.1) and (1.2), parameter r1>0 is the intrinsic growth rate of species x1, ri≤0 represents the death rate of species xi for i=2,3,⋯,n, hi≥0 represents the harvesting rate of human for the species xi for i=1,2,⋯,n, xi(t−) is the left limit of xi(t) at time t, aii>0(i=1,2,⋯,n) is the density dependent coefficient of species xi, aii+1≥0(i=2,3,⋯,n−1) is the capture rate of which species xi+1 preys on species xi, aii−1≥0(i=2,3,⋯,n) stands for efficiency of food conversion from species xi−1 to species xi. Bi(t)(i=1,2,⋯,n) are independent standard Brownian motion, and σ2i>0 is the intensity of Bi(t). Lévy noises ˜N(dt,du)=N(dt,du)−λ(du)dt, where N(dt,du) is a poisson counting measure with characteristic measure λ on a Borel-measurable subset Z of (0,∞) satisfying λ(Z)<+∞. Functions γi(u):Z→R(i=1,2,⋯,n) represent the effects of random jumps for species xi that are assumed to be continuous on Z. In this paper we further assume that Lévy noises ˜N(dt,du) and Brownian motion Bi(t)(i=1,2,⋯,n) are also independent.
For any solution (x1(t),⋯,xn(t)) of model (1.2), the initial value is given as follows.
xi(0)=xi0,i=1,2,⋯,n, | (2.1) |
where xi0(i=1,2,⋯,n) are positive constants.
Denote bi=ri+∫z[ln(1+γi(u))−γi(u)]λ(du)−σ2i2 and ci=bi−hi for i=1,2,⋯,n, and αk=(c1,c2,⋯,ck)T for k=1,2,⋯,n. Furthermore, we denote
Ak=[a11a120⋯000−a21a22a23⋯000⋯⋯⋯⋯⋯⋯⋯000⋯−ak−1k−2ak−1k−1ak−1k000⋯0−akk−1akk]. |
Particularly, we have
A1=[a11],A2=[a11a12−a21a22]. |
Denote by Hk the determinant of Ak. Let A(i)k be the matrix gained from matrix Ak by changing its i-th column to αk, and D(i)k is the determinant of A(i)k. Furthermore, for any square matrix P we denote by P(i,j) the algebraic complement corresponding to the element of i-th row and j-th column of P.
For the Lévy noises in models (1.1) and (1.2), we always assume that the following condition is satisfied.
(H) For i=1,2,⋯,n and u∈Z, 1+γi(u)>0 holds, and a constant z>0 exists such that
∫Z[ln(1+γi(u))]2λ(du)<z. |
Using the same method as in [20], the following conclusions can be established for model (1.2).
Lemma 1. Hk>0 for every k=1,2,⋯,n.
Lemma 2. For any initial values (x10,x20,⋯,xn0)∈Rn+, model (1.2) possesses a globally unique positive solution x(t)=(x1(t),⋯,xn(t))∈Rn+a.s. for all t≥0 satisfying the initial condition (2.1). Furthermore, for any constant q>0, there are constants Ci(q)>0(i=1,2,⋯,n) such that
lim supt→∞E[xqi(t)]≤Ci(q),i=1,2,⋯,n. |
Similar to [20], we define
N1=a12a11A(2)k(3,2)Hk(3,k),Ni=aii−1aiiNi−1+aii+1aiiA(i+1)k(i+2,i+1)Hk(i+2,k),i=1,2,⋯,k−2,Nk−1=ak−1k−2ak−1k−1Nk−2. |
Lemma 3. For any positive solution (x1(t),⋯,xn(t)) of model (1.2), assume that D(n)n>0 and (1+ann−1annNn−1)>0 if n≥3, then limt→∞⟨xi(t)⟩=D(i)nHn a.s., i=1,2,⋯,n.
A simple calculation we can obtain that for any integers k,i,j∈{1,2,⋯,n} with i,j≤k
A(i)k(j,i)=Ak(j,i). |
Therefore, we further have
N1=a12a11Ak(3,2)Hk(3,k),Ni=aii−1aiiNi−1+aii+1aiiAk(i+2,i+1)Hk(i+2,k),i=1,2,⋯,k−2,Nk−1=ak−1k−2ak−1k−1Nk−2. | (2.2) |
Theorem 1. Let (x1(t;ψ),x2(t;ψ),⋯,xn(t;ψ)) and (z1(t;ψ∗),z2(t;ψ∗),⋯,zn(t;ψ∗)) be two solutions of model (1.2) with initial values ψ,ψ∗∈Rn+. Assume that there are positive constants wi(i=1,2,⋯,n) such that
w1a11−w2a21>0,wiaii−wi−1ai−1i−wi+1ai+1i>0(i=2,3,⋯,n−1),wnann−wn−1an−1n>0. |
Then we have
limt→∞E(n∑i=1|xi(t,ψ)−zi(t,ψ∗)|2)12=0. |
Proof. We only need to show
limt→∞E|xi(t;ψ)−zi(t;ψ∗)|=0,i=1,2,⋯,n. | (3.1) |
Define the functions as follows
Vi(xi)=|lnxi(t;ψ)−lnzi(t;ψ∗)|,i=1,2,⋯,n. |
Applying Itô formula, then
LVi(xi)≤−aii|xi(t;ψ)−zi(t;ψ∗)|+aii−1|xi−1(t;ψ)−zi−1(t;ψ∗)|+aii+1|xi+1(t;ψ)−zi+1(t;ψ∗)|,i=1,2,⋯,n. | (3.2) |
where a10=ann+1=0 are stupulated. Define the following function
V(t)=n∑i=1wiVi(xi). | (3.3) |
From (3.1) we obtain
LV(t)=n∑i=1wiLVi(xi)≤−n∑i=1(wiaii−wi−1ai−1i−wi+1ai+1i)|xi(t;ψ)−zi(t;ψ∗)|, |
where a01=an+1n=0 and ω0=ωn+1=0 are stipulated. Therefore,
E[V(t)]≤E[V(0)]−n∑i=1(wiaii−wi−1ai1i−wi+1ai+1i)∫t0E[|xi(s;ψ)−zi(s;ψ∗)]|ds. |
which implies
∫∞0E[|xi(s;ψ)−zi(s;ψ∗)|]ds<+∞,i=1,2,⋯,n. | (3.4) |
Define the functions
Fi(t)=E[|xi(t;ψ)−zi(t;ψ∗)|],i=1,2,⋯,n. |
Then, for any t1,t2∈[0,+∞) and each i=1,2,⋯,n,
|Fi(t2)−Fi(t1)|=|E[|xi(t2;ψ)−zi(t2;ψ∗)|−|xi(t1;ψ)−zi(t1;ψ∗)|]|≤E[|xi(t2;ψ)−xi(t1;ψ)|]+E[|zi(t2;ψ∗)−zi(t1;ψ∗)|]. | (3.5) |
Applying Itô formula to model (1.2), then
xi(t2;ϕ)−xi(t1;ϕ)=∫t2t1xi(s;ϕ)[ri−hi+aii−1xi−1(s;ϕ)−aiixi(s)−aii+1xi+1(s;ϕ)]ds+∫t2t1σixi(s;ϕ)dBi(s)+∫t2t1∫Zxi(s−)γi(u)˜N(ds,du),i=1,2,...,n. | (3.6) |
For any t2>t1 and q>1, from the first equation of (3.6) and Hölder's inequality, then
(E[|x1(t2;ψ)−x1(t1;ψ)|])q≤E[(∫t2t1x1(s;ψ)|r1−h1−a11x1(s;ψ)−a12x2(s;ψ)|ds+|∫t2t1σ1x1(s;ψ)dB1(s)|)q]+|∫t2t1∫Zx1(s−)γ1(u)˜N(ds,du)|≤3qE[(∫t2t1x1(s;ψ)|r1−h1−a11x1(s;ψ)−a12x2(s;ψ)|ds)q]+3qE[|∫t2t1σ1x1(s;ψ)dB1(s)|q]+3qE[|∫t2t1∫Zxi(s−)γi(u)˜N(ds,du)|q]. | (3.7) |
Similarly,
E[(∫t2t1x1(s;ψ)|r1−h1−a11x1(s;ψ)−a12x2(s;ψ)|ds)q]≤(t2−t1)q−1E[∫t2t1(|r1−h1|x1(s;ψ)+a11x21(s;ψ)+a12x1(s;ψ)x2(s;ψ))qds]≤3q(t2−t1)q−1|r1−h1|q∫t2t1E[xq1(s;ψ)]ds+3qaq11(t2−t1)q−1∫t2t1E[x2q1(s;ψ)]ds+3q(t2−t1)q−1E[∫t2t1(a12x1(s;ψ)x2(s;ψ))qds]. | (3.8) |
and
E[∫t2t1(a12x1(s;ψ)x2(s;ψ))qds]≤E[∫t2t1(12a12x21(s;ψ)+12a12x22(s;ψ))qds]≤aq12∫t2t1E[x2q1(s;ψ)]ds+aq12∫t2t1E[x2q2(s;ψ)]ds. | (3.9) |
In view of Theorem 7.1 in [21] on the Burkholder-Davis-Gundy inequality, for 1<q≤2 and any t2>t1, then
E[|∫t2t1σ1x1(s;ψ)dB1(s)|q]≤DqE[∫t2t1|σ1x1(s;ψ)|2ds]q/2≤Dqσq1(t2−t1)q−22∫t2t1E[xq1(s;ψ)]ds. | (3.10) |
Using the Kunita first inequality in [21], there exists a constant Kq>0 and from assumption (H) we can obtain
E[|∫t2t1∫Zx1(s;ψ)γ1(u)˜N(ds,du)|q]≤KqE[∫t2t1∫Z|x1(s)γ1(u)|2λ(du)ds]q/2+E[∫t2t1∫Z|x1(s)γ1(u)|qλ(du)ds]≤KqLq/2(t2−t1)q−22∫t2t1E[xq1(s;ψ)]ds+KqLq/2∫t2t1E[xq1(s;ψ)]ds. | (3.11) |
From Lemma 2, there exists C∗∗i(q)>0 for i=1,2,⋯,n such that supt≥−γE[xqi(t)]≤C∗∗i(q). Therefore, from (3.7)–(3.10) there is a δ>0 satisfies that for 1<q≤2 and any t1≥0,t2≥0 with |t2−t1|≤δ, such that
(E[|x1(t2;ψ)−x1(t1;ψ)|])q≤3q[Dqσq1(t2−t1)q2C∗∗1(q)]+(t2−t1)q23q[KqLq/2C∗∗1(q)+KqLq/2(t2−t1)1−q/2C∗∗1(q)]+3q[3q(t2−t1)q|r1−h1|qC∗∗1(q)+3qaq11(t2−t1)qC∗∗1(2q)]+32q(t2−t1)qaq12[C∗∗1(2q)+C∗∗2(2q)]≤M∗∗1|t2−t1|q2, |
where
M∗∗1=3qDqσq1C∗∗1(q)+3qKqLq/2[1+δ2−q2]C∗∗1(q)+[81δ]q2[|r1−h1|qC∗∗1(q)+aq11C∗∗1(2q)]+[81δ]q2aq12[C∗∗1(2q)+C∗∗2(2q)]. |
Similarly, we also obtain
(E[|z1(t2;ψ∗)−z1(t1;ψ∗)|])q≤M∗∗1|t2−t1|q2. |
for any t1≥0,t2≥0 with |t2−t1|≤δ and 1<q≤2. Thus, from (3.5), we obtain
|F1(t2)−F1(t1)|≤E[|x1(t2;ψ)−x1(t1;ψ)|]+E[|z1(t2;ψ∗)−z1(t1;ψ∗)|]≤2(M∗∗1)1q√|t2−t1|. | (3.12) |
Using similar argument, for Fi(t)(i=2,3,⋯,n), we also can obtain that there is a δ>0 for any t1≥0,t2≥0 with |t2−t1|≤δ and 1<q≤2
|Fi(t2)−Fi(t1)|≤2(M∗∗i)1q√|t2−t1|,i=2,3,⋯,n, | (3.13) |
where
M∗∗i=3qDqσqiC∗∗i(q)+3qKqLq/2[1+δ2−q2]C∗∗i(q)+[144δ]q2[|ri+hi|qC∗∗i(q)+aqiiC∗∗i(2q)]+[144δ]q2aqii+1[C∗∗i(2q)+C∗∗i+1(2q)]+[144δ]q2aqii−1[C∗∗i−1(2q)+C∗∗i+1(2q)],i=2,3,⋯,n. |
From (3.12) and (3.13), we obtain that Fi(t),i=1,2,⋯,n for t∈(0,∞) are uniformly continuous. Therefore, from (3.4) and Barbalat lemma in [22] we finally obtain that (3.1) holds. This completes the proof.
Let P(Rn+) be the space of all probability measures on C(Rn+). For Q1,Q2∈P(Rn+), we define
dLB(Q1,Q2)=supf∈LB|∫Rn+f(y)Q1(dy)−∫Rn+f(y)Q2(dy)|, | (3.14) |
where
LB={f:Rn+→R:|f(y1)−f(y2)|≤‖y1−y2‖,|f(⋅)|≤1}. | (3.15) |
Denote by p(t,ϕ,dx) the transition probability of process x(t)=(x1(t),x2(t),⋯,xn(t)). We have the following results.
Theorem 2. Assume that there exists constants wi>0(i=1,2,⋯,n) such that
w1a11−w2a21>0,wiaii−wi−1ai−1i−wi+1ai+1i>0(i=2,3,⋯,n−1),wnann−wn−1an−1n>0. |
Then model (1.2) is asymptotically stable in distribution, that is, there is a probability measure v(⋅) satisfies that for any initial function ψ∈Rn+, the transition probability p(t,ψ,⋅) of solution (x1(t,ψ),x2(t,ψ),⋯,xn(t,ψ)) such that
limt→∞dLB(p(t,ψ,⋅),v(⋅))=0. |
Proof. For any initial data ψ∈Rn+, the family probability p(t,ψ,⋅) is tight which obtained from Theorem 1 and Chebyshev's inequality. For any f∈LB and t,s>0, we compute
∣Ef(y(t+s;ψ))p(t+s,ψ,dy)−Ef(y(t;ψ))p(t,ψ,dy)∣=∣E[E(f(y(t+s;ψ))∣Fs)]−E[f(y(t;ψ))]∣=∣∫Rn+E[f(y(t;ϕ))]p(s,ψ,dϕ)−E[f(y(t;ψ))]∣≤∫Rn+∣Ef(y(t;ϕ))−Ef(y(t;ψ))∣p(s,ψ,dϕ). |
It is clear that there exists a constant T>0 from (3.1), such that for all t≥T,
supf∈LB∣Ef(y(t;ϕ))−Ef(y(t;ψ))∣≤ε. |
Consequently
∣Ef(y(t+s;ψ))−Ef(y(t;ψ))∣≤ε. |
Since f is arbitrary, we gain
supf∈LB∣Ef(y(t+s;ψ))−Ef(y(t;ψ))∣≤ε. |
That is,
dLB(p(t+s,ψ,⋅),p(t,ψ,⋅))≤ε. |
Therefore {p(t,ψ,⋅):t≥0} is Cauchy in the space P(Rn+). Hence, there is a unique v(⋅)∈P(Rn+) such that
limt→∞dLB(p(t,ψ0,⋅),v(⋅))=0. |
where ψ0=(c10,c20,...,cn0)T∈Rn+, and c10,⋯,cn0 are positive constants. It then follows from (3.1) that
limt→∞dLB(p(t,ψ,⋅),p(t,ψ0,⋅)=0. |
Hence,
limt→∞dLB(p(t,ψ,⋅),v(⋅)≤limt→∞dLB(p(t,ψ,⋅),p(t,ψ0,⋅))+limt→∞dLB(p(t,ψ0,⋅),v(⋅))=0. |
This completes the proof.
Since Hn>0 in Lemma 1, the inverse matrix A−1n exists. Let B=(b1,b2,⋯,bn)T and H=(h1,h2,⋯,hn)T. Moreover, let E be the unit matrix, and H∗=(An(A−1n)T+E)−1B≜(h∗1,h∗2,⋯,h∗n)T.
Theorem 3. Assume that the following conditions hold
(1) (1+ann−1annNn−1)>0 if n≥3;
(2) there exists constants wi>0(i=1,2,⋯,n) such that
w1a11−w2a21>0,wiaii−wi−1ai−1i−wi+1ai+1i>0(i=2,3,⋯,n−1),wnann−wn−1an−1n>0. |
Then the following conclusions hold.
(A1) If D(n)n|hi=h∗i>0, A−1n+(A−1n)T is positive semi-definite and h∗i≥0 (i=1,2,⋯,n), then model (1.2) has the OHS H=H∗, and
MESY≜S(H∗)=(H∗)TA−1n(B−H∗). | (4.1) |
(A2) If any of the following conditions stands, then the OHS for model (1.2) does not exist.
(B1) D(n)n|hi=h∗i,i=1,2,⋯,n≤0;
(B2) There is a h∗i with h∗i<0;
(B3) A−1n+(A−1n)T is not positive semi-definite.
Proof. Let Y={H=(h1,h2,⋯,hn)T∈Rn:D(n)n>0,hi≥0,i=1,2,⋯,n}, which means that for any H∈Y, the conclusion of Lemma 3 stands.
First, we prove the conclusion (A1). Since H∗∈Y, the set Y is not empty. A unique probability measure v(⋅) is gained from Theorem 2. Then, v(⋅) is also strong mixing by Corollary 3.4.3 [23]. From Theorem 3.2.6[23], we gain that the measure v(⋅) is ergodic. For any initial value ψ∈Rn+, model (1.2) possess a positive gloabl solution x(t)=(x1(t),x2(t),⋯,xn(t)). It yields by Theorem 3.3.1 [23] that
limt→∞1t∫t0HTx(s)ds=∫Rn+HTxv(dx), | (4.2) |
where H=(h1,h2,⋯,hn)T∈Y. Let ϱ(z) be the stationary probability density of model (1.2), then we have
S(H)=limt→∞E[n∑i=1hixi(t)]=limt→∞E[HTx(t)]=∫Rn+HTxϱ(x)dx. | (4.3) |
In view of the invariant measure is sole, there also exists a one-to-one correspondence among ϱ(z) and its corresponding invariant measure. Hence, we have
∫Rn+HTxϱ(x)dx=∫Rn+HTxv(dx). | (4.4) |
Combining (4.2)–(4.4) with the conclusion of Lemma 3, then
S(H)=limt→+∞1t∫t0HTx(s)ds=h1limt→+∞1t∫t0x1(s)ds+h2limt→+∞1t∫t0x2(s)ds+⋯+hnlimt→+∞1t∫t0xn(s)ds=h1D(1)nHn+h2D(2)nHn+⋯+hnD(n)nHn. |
Moreover, S(H)=HT(An)−1(B−H) is derived by careful calculation. Computing the gradient of S(H), then
∂S(H)∂H=∂HT∂H(An)−1(B−H)+∂(B−H)T∂H((An)−1)TH. |
Since ∂HT∂H=E, then
∂S(H)∂H=(An)−1(B−H)−((An)−1)TH=(An)−1B−((An)−1+((An)−1)T)H. |
By solving ∂S(H)∂H=0, the following critical value is gained
H=((An)−1+((An)−1)T)−1(An)−1B. |
That is,
H=[(An)−1((An)((An)−1)T+E)]−1(An)−1B=((An)((An)−1)T+E)−1B=H∗. |
Calculating the Hessian matrix of S(H), we obtain
∂∂H(∂S(H)∂H)=−∂((An)−1+((An))T)H∂H=−((An)−1+((An)−1)T). | (4.5) |
Since (An)−1+((An)−1)T is positive semi-definit, from the existence principle of extremum value of multi-variable function we obtain that S(H) has a global maximum value at H=H∗. Obviously, H∗ is also sole. Therefore, if H∗∈Y, that is h∗i≥0(i=1,2,⋯,n) and D(n)n|hi=h∗i,i=1,2,⋯,n>0, then we finally obtain that H∗ is an OHS, and MESY shown in (20).
Next, we prove the conclusion (A2). We first assume that (B1) or (B2) stands. Suppose that ˜Γ=(γ1,γ2,⋯,γn) is the OHS for model (1.2), thus Γ∈Y, that is
D(n)n|hi=γi,i=1,2,⋯,n>0,γi≥0,i=1,2,⋯,n. | (4.6) |
On the other hand, if Γ∈Y is the OHS, then Γ is a unique solution of equation ∂S(H)∂H=0. Therefore, we have (h∗1,h∗2,⋯,h∗n)=(γ1,γ2,⋯,γn). Thus, condition (4.6) becomes into
D(n)n|hi=h∗i,i=1,2,⋯,n>0,h∗i≥0,i=1,2,⋯,n. |
However, this is impossible.
Finally, consider (B3). It can be assumed that (B1) and (B2) fail to stand in this case. Therefore, h∗i≥0(i=1,2,⋯,n) and D(n)n|hi=h∗i,i=1,2,⋯,n>0. Thus, Y is not empty, i.e., (4.2)–(4.4) hold. Denote (An)−1+((An)−1)T=(hij)n×n. Then, by calculating we have
h11=2(a22a33⋯ann+⋯+a22a34⋯ann−1+a23a32⋯ann)Hn |
Obviously, h11>0, which means that (An)−1+((An)−1)T is not negative semi-definite. It can be further derived from the conditions in (B3) that (An)−1+((An)−1)T is indefinite. Hence, there is not the OHS if B3 holds. This completes the proof.
Remark 1. In Theorem 3 we see that the conditions (1) and (2) are required. From the following argumentations we will find that when n=3 and n=4 condition (2) can bring out condition (1). Therefore, we propose the following proposition. However, when n>4 whether condition (2) can bring out condition (1) still is an open problem. Thus, the following conjecture can be proposed.
Proposition 1. When n=3 and n=4, if condition (2) holds then condition (1) also holds.
Proof. Let k=3, then condition (2) becomes to
w1a11−w2a21>0,w2a22−w1a12−w3a32>0w3a33−w2a23>0. | (4.7) |
From (2.2), by calculating we have
N1=−a12a23a11a22+a12a21,N2=−a12a21a23a22(a11a22+a12a21), |
and
1+a32a33N2=a22a33(a11a22+a12a21)−a12a21a23a32a22a33(a11a22+a12a21). |
From (4.7) we derive
a22≥1w2(w1a12)+w3a32,a33≥w2a23w3. |
Therefore, we further have
a22a33(a11a22+a12a21)−a12a21a23a32>[1w2(w1a12+w3a32)1w3w2a23(a11a22+a12a21)]−a12a21a23a32=(w1a12w2w3+a32w2)w2a23(a11a22+a12a21)−a12a21a23a32=(w1a12a23w3+a23a32)(a11a22+a12a21)−a12a21a23a32=w1a12a23w3(a11a22+a12a21)+a11a22a23a32>0. |
This shows that 1+a32a33N2>0.
Let k=4, then condition (2) becomes to
w1a11−w2a21>0,w2a22−w1a12−w3a32>0,w3a33−w2a23−w4a43>0,w4a44−w3a34>0. | (4.8) |
From (2.2), by calculating, we have
N1=−a12a23a44a43H2,N2=−(a12a21a23a44H3+a23a34a43H22)a22a43H2H3,N3=−a32(a12a21a23a44H3+a23a34a43H22)a33a22a43H2H3, |
and
1+a43a44N3=1−(a12a21a23a32a44H3+a23a32a34a43H22)a22a33a44H2H3=a22a33a44H2H3−a12a21a23a32a44H3−a23a32a34a43H22a22a33a44H2H3. |
From (4.8) we obtain
a11>w2w1a21,aii>1wi(wi−1ai−1i+wi+1ai+1i),i=2,3,a44>w3w4a34. |
Therefore,
(a22a33H2−a12a21a23a32)a44H3>(1w2[w1a12+w3a32]1w3[w2a23+w4a43][a11a22+a12a21]−a12a21a23a32)a44H3≥((1w2w1a121w3[w2a23+w4a43]+w4w2a32a43)(a11a22+a12a21))a44H3≥w4w2a32a43(a11a22+a12a21)w3w4a34H3. |
Since
H3>(a11a22+a12a21)a33>(a11a22+a12a21)w2w3a23, |
we further obtain
(a22a33H2−a12a21a23a32)a44H3>w4w2a32a43(a11a22+a12a21)w3w4a34(a11a22+a12a21)w2w3a23=a23a32a34a43H22. |
This shows that 1+a43a44N3>0.
Conjecture 1. When n>4, if condition (2) holds then condition (1) also holds.
To support our analysis results, we will provide the numerical examples for model (1.2). The numerical simulations can be performed qualitatively, and parameters are selected in an epidemiologically feasible way.
Example 1. In model (1.2) we take n=4, r1=1.5, r2=−0.6, r3=−0.08, r4=−0.01, γ1=0.1, γ2=0.15, γ3=0.15, γ4=0.1, a11=1.6, a12=0.2, a22=2.2, a21=2.5, a23=0.1, a32=2.5, a33=2, a34=0.3, a43=1.3, a44=2, σ1=0.1, σ2=0.2, σ3=0.1 and σ4=0.1
Choose w1=2.5, w2=1.3, w3=0.8 and w4=0.5, by calculating we have w1a11−w2a21=0.7500>0, w2a22−w1a12−w3a32=0.3600>0, w3a33−w2a23−w4a43=0.8200>0 and m4a44−m3a34=0.7600>0. Proposition 1 shows that 1+a43a44N3>0. Thus, all four species in model (1.2) are persistent in mean which is shown in Figure 1 below. Furthermore, h∗1=0.3250, h∗2=0.3914, h∗3=0.3264, h∗4=0.1281 and D(4)4|hi=h∗i,i=1,2,3,4=0.5201>0 are obtained by calculating, all leading principal minors of A4+(A−14)T are positive, which shows that the A4+(A−14)T is positive semi-definite. Thus, the condition of conclusion (A1) is satisfied, then the OHS: H∗=(0.3250, 0.3914, 0.3264, 0.1281)T and MESY=0.4321 are gained, as Figure 1 shows below.
Example 2. In model (1.2) we take n=4, r1=1.5, r2=−0.5, r3=−0.03, r4=−0.01; γ1=0.15, γ2=0.2, γ3=0.1 and γ4=0.05. The rest parameters in model (1.2) are shown in the following three cases.
Case 2.1. a11=1.6, a12=0.2, a22=2, a21=2.5, a23=1, a32=2.5, a33=2.1, a34=0.2, a43=0.1, a44=2, σ1=0.1, σ2=0.2, σ3=0.2 and σ4=0.1.
Case 2.2. a11=1.8, a12=0.2, a22=2, a21=1.5, a23=0.2, a32=1.5, a33=2.1, a34=0.8, a43=1.1, a44=2.5, σ1=0.2, σ2=0.2, σ3=0.3 and σ4=0.1.
Case 2.3. a11=0.8, a12=0.2, a22=2, a21=1, a23=0.15, a32=1.5, a33=2, a34=0.1, a43=1, a44=0.08, σ1=0.2, σ2=0.3, σ3=0.3 and σ4=0.1.
For Case 2.1, choose w1=2.5, w2=1.3, w3=0.8 and w4=1.4, then we have w1a11−w2a21=0.75>0, w2a22−w1a12−w3a32=0.10>0, w3a33−w2a23−w4a43=0.24>0 ºÍ m4a44−m3a34=2.64>0. Furthermore, h∗1=0.3621, h∗2=0.4773, h∗3=0.4665, h∗4=0.0051 are gained by calculating. Meanwhile, D(4)4|hi=h∗i,i=1,2,3,4=−0.1475<0. Thus the condition (B1) of (A2) is satisfied. Therefore, the OHS do not exist. The numerical simulations are given in Figure 2.
For Case 2.2, choose w1=2.2, w2=1.1, w3=1.1 and w4=1.4, then we have w1a11−w2a21=2.31>0, w2a22−w1a12−w3a32=0.11>0, w3a33−w2a23−w4a43=0.55>0 ºÍ m4a44−m3a34=2.62>0. Furthermore, h∗1=0.9445>0, h_2^* = 0.2174 > 0 , h_3^* = 0.0039 > 0 and h_4^* = -0.0124 < 0 are obtained by calculating. Thus the condition ( \mathcal{B}_2 ) of conclusion ( \mathcal{A}_2 ) is satisfied. Therefore, the OHS do not exist. The numerical simulations are given in Figure 3.
For Case 2.3, choose w_1 = 2.8 , w_2 = 2 , w_3 = 2 and w_4 = 2.8 , then we have w_1a_{11}-w_2a_{21} = 0.24 > 0 , w_2a_{22}-w_1a_{12}-w_3a_{32} = 0.44 > 0 , w_3a_{33}-w_2a_{23}-w_4a_{43} = 0.90 > 0 ºÍ m_4a_{44}-m_3a_{34} = 0.024 > 0 . Furthermore, by calculating we obtain that the fourth leading principle minor of A_4+(A_4^{-1})^T is equal to -6.2127 < 0 , which means that A_4+(A_4^{-1})^T is negative semi-definite. Thus the condition ( \mathcal{B}_3 ) is satisfied. Therefore, the optimal harvesting strategy do not exist. The numerical simulations are given in Figure 4.
In this paper, based on the effect of harvesting and Lévy noises, a stochastic marine n -species system is investigated. On the basis of the research results in [23], we first establish the criterion for the stability in distribution of model (1.2). Then, the criteria for existence and nonexistence of OHS and the corresponding MESY are derived. This shows that the environmental noises affect the OHS and MESY.
Compared with the existing results, our first contribution is to investigate the stochastic stability and optimal harvesting problems for a stochastic marine n -species food chain model with Lévy noises and white noises. Second, the criteria on the stability in distribution and optimal harvesting strategy for stochastic n -species food chain model are established for the first time. Third, the noises intensity can easily affect the dynamics and MESY of marine populations, and further lead to the imbalances of marine ecology. Finally, we find that for better manage and protect marine resources, the establishment of an optimal harvesting strategy should fully consider the impact of noise intensity.
There are still some remained problems that need to be followed up. For example, in Conjecture 1, we assume if w_1a_{11}-w_2a_{21} > 0\text{, } \; w_ia_{ii}-w_{i-1}a_{i-1i}-w_{i+1}a_{i+1i} > 0\; (i = 2\text{, }3\text{, }\cdots\text{, } n-1)\text{, }\; w_na_{nn}-w_{n-1}a_{n-1n} > 0 , then 1+\frac{a_{nn-1}}{a_{nn}}N_{n-1} > 0 , how to prove is an open problem. Second, marine population are easily affected by climate change, consider marine population model with regime switching is necessary. Finally, more complex and realistic stochastic system with nonlinear functional response, delays and Allee effect need to be investigated. These issues may be investigated in the future.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
This work was supported by the National Natural Science Foundation of China (Grant No. 12061079), Foundation of Fujian Educational Committee (Grant No. JAT220189), Startup Foundation of Jimei University(Grant No.ZQ2022025).
The authors declare there is no conflicts of interest.
[1] |
J. Melnikow, A. Padovani, M. Miller, Frontline physician burnout during the COVID-19 pandemic: national survey findings, BMC Health Serv. Res., 22 (2022), 365. https://doi.org/10.1186/s12913-022-07728-6 doi: 10.1186/s12913-022-07728-6
![]() |
[2] |
J. Mapes, Using big data to study small places: small-town voting patterns in the 2020 US presidential election, Growth Change, 55 (2024), e12730. https://doi.org/10.1111/grow.12730 doi: 10.1111/grow.12730
![]() |
[3] |
E. Clough, T. Barrett, S. E. Wilhite, P. Ledoux, C. Evangelista, I. F. Kim, et al., NCBI GEO: archive for gene expression and epigenomics data sets: 23-year update, Nucleic Acids Res., 52 (2024), D138–D144. https://doi.org/10.1093/nar/gkad965 doi: 10.1093/nar/gkad965
![]() |
[4] |
S. Yang, D. Meng, A. Díaz, H. Yang, X. Su, A. M. D. Jesus, Probabilistic modeling of uncertainties in reliability analysis of mid-and high-strength steel pipelines under hydrogen-induced damage, Int. J. Struct. Integr., 16 (2025), 39–59. https://doi.org/10.1108/IJSI-10-2024-0177 doi: 10.1108/IJSI-10-2024-0177
![]() |
[5] | S. Yang, D. Meng, H. Yang, C. Luo, X. Su, Enhanced soft Monte Carlo simulation coupled with support vector regression for structural reliability analysis, In: Proceedings of the Institution of Civil Engineers-Transport, Emerald Publishing Limited, 2024, 1–16. https://doi.org/10.1680/jtran.24.00128 |
[6] |
D. Meng, H. Yang, S. Yang, Y. Zhang, A. M. De Jesus, J. Correia, et al., Kriging-assisted hybrid reliability design and optimization of offshore wind turbine support structure based on a portfolio allocation strategy, Ocean Eng., 295 (2024), 116842. https://doi.org/10.1016/j.oceaneng.2024.116842 doi: 10.1016/j.oceaneng.2024.116842
![]() |
[7] | S. Sen, S. K. Ghosh, H. Al-Mofleh, The Mirra distribution for modeling time-to-event data sets, In: Dtrategic management, decision theory, and decision science: contributions to policy issues, Singapore: Springer, 2021. https://doi.org/10.1007/978-981-16-1368-5_5 |
[8] |
D. Roy, Discrete Rayleigh distribution, IEEE Trans. Reliab., 53 (2004), 255–260. https://doi.org/10.1109/TR.2004.829161 doi: 10.1109/TR.2004.829161
![]() |
[9] |
M. El-Morshedy, M. S. Eliwa, E. Altun, Discrete Burr-Hatke distribution with properties, estimation methods and regression model, IEEE Access, 8 (2020), 74359–74370. https://doi.org/10.1109/ACCESS.2020.2988431 doi: 10.1109/ACCESS.2020.2988431
![]() |
[10] |
H. Krishna, P. S. Pundir, Discrete Burr and discrete Pareto distributions, Stat. Methodol., 6 (2009), 177–188. https://doi.org/10.1016/j.stamet.2008.07.001 doi: 10.1016/j.stamet.2008.07.001
![]() |
[11] | T. Hussain, M. Ahmad, Discrete inverse Rayleigh distribution, Pakistan J. Stat., 30 (2014). |
[12] | A. E. Abd EL-Hady, M. A. Hegazy, A. A. EL-Helbawy, A discrete exponentiated generalized family of distributions, Comput. J. Math. Stat. Sci., 2 (2023), 303–327. |
[13] | A. R. E. Alosey, A. M. Gemeay, A novel version of geometric distribution: method and application, Comput. J. Math. Stat. Scie., 4 (2025), 1–16. |
[14] |
M. A. Jazi, C. D. Lai, M. H. Alamatsaz, A discrete inverse Weibull distribution and estimation of its parameters, Stat. Methodol., 7 (2010), 121–132. https://doi.org/10.1016/j.stamet.2009.11.001 doi: 10.1016/j.stamet.2009.11.001
![]() |
[15] |
E. Gómez-Déniz, E. Calderín-Ojeda, The discrete Lindley distribution: properties and applications, J. Stat. Comput. Simul., 81 (2011), 1405–1416. https://doi.org/10.1080/00949655.2010.487825 doi: 10.1080/00949655.2010.487825
![]() |
[16] |
J. M. Jia, Z. Z. Yan, X. Y. Peng, A new discrete extended Weibull distribution, IEEE Access, 7 (2019), 175474–175486. https://doi.org/10.1109/ACCESS.2019.2957788 doi: 10.1109/ACCESS.2019.2957788
![]() |
[17] |
E. Altun, A new generalization of geometric distribution with properties and applications, Commun. Stat.-Simul. Comput., 49 (2020), 793–807. https://doi.org/10.1080/03610918.2019.1639739 doi: 10.1080/03610918.2019.1639739
![]() |
[18] |
E. Gómez-Déniz, Another generalization of the geometric distribution, Test, 19 (2010), 399–415. https://doi.org/10.1007/s11749-009-0169-3 doi: 10.1007/s11749-009-0169-3
![]() |
[19] | M. A. Hegazy, R. E. Abd El-Kader, A. A. El-Helbawy, G. R. Al-Dayian, Bayesian estimation and prediction of discrete Gompertz distribution, J. Adv. Math. Comput. Sci., 36 (2021), 1–21. |
[20] |
V. Nekoukhou, M. H. Alamatsaz, H. Bidram, Discrete generalized exponential distribution of a second type, Statistics, 47 (2013), 876–887. https://doi.org/10.1080/02331888.2011.633707 doi: 10.1080/02331888.2011.633707
![]() |
[21] |
A. S. Eldeeb, M. Ahsan-ul-Haq, M. S. Eliwa, A discrete Ramos-Louzada distribution for asymmetric and over-dispersed data with leptokurtic-shaped: properties and various estimation techniques with inference, AIMS Math., 7 (2022), 1726–1741. https://doi.org/10.3934/math.2022099 doi: 10.3934/math.2022099
![]() |
[22] | T. Hussain, M. Aslam, M. Ahmad, A two parameter discrete Lindley distribution, Rev. Colomb. Estad., 39 (2016), 45–61. |
[23] | B. C. Arnold, N. Balakrishnan, H. N. Nagaraja, A first course in order statistics, Society for Industrial and Applied Mathematics, 2008. |
[24] |
R. Shanker, H. Fesshaye, On Poisson-Sujatha distribution and its applications to model count data from biological sciences, Biometrics Biostat. Int. J., 3 (2016), 1–7. https://doi.org/10.15406/bbij.2015.02.00036 doi: 10.15406/bbij.2015.02.00036
![]() |
[25] | D. J. Hand, F. Daly, K. McConway, D. Lunn, E. Ostrowski, A handbook of small data sets, Chapman & Hall, 1993. https://doi.org/10.1201/9780429246579 |
1. | A. M. Elaiw, A. D. Al Agha, S. A. Azoz, E. Ramadan, Global analysis of within-host SARS-CoV-2/HIV coinfection model with latency, 2022, 137, 2190-5444, 10.1140/epjp/s13360-022-02387-2 | |
2. | N. H. AlShamrani, Stability of an HTLV-HIV coinfection model with multiple delays and CTL-mediated immunity, 2021, 2021, 1687-1847, 10.1186/s13662-021-03416-7 | |
3. | M.A. Alshaikh, N.H. AlShamrani, A.M. Elaiw, Stability of HIV/HTLV co-infection model with effective HIV-specific antibody immune response, 2021, 27, 22113797, 104448, 10.1016/j.rinp.2021.104448 | |
4. | M. Pradeesh, A. Manivannan, S. Lakshmanan, F. A. Rihan, Prakash Mani, Abdellatif Ben Makhlouf, Dynamical Analysis of Posttreatment HIV-1 Infection Model, 2022, 2022, 1099-0526, 1, 10.1155/2022/9752628 | |
5. | Muhammad Shoaib, Nabeela Anwar, Iftikhar Ahmad, Shafaq Naz, Adiqa Kausar Kiani, Muhammad Asif Zahoor Raja, Intelligent networks knacks for numerical treatment of nonlinear multi-delays SVEIR epidemic systems with vaccination, 2022, 36, 0217-9792, 10.1142/S0217979222501004 | |
6. | Shiv Mangal, O.P. Misra, Joydip Dhar, Fractional-order deterministic epidemic model for the spread and control of HIV/AIDS with special reference to Mexico and India, 2023, 03784754, 10.1016/j.matcom.2023.03.008 | |
7. | Noura H. AlShamrani, Ahmed Elaiw, Aeshah A. Raezah, Khalid Hattaf, Global Dynamics of a Diffusive Within-Host HTLV/HIV Co-Infection Model with Latency, 2023, 11, 2227-7390, 1523, 10.3390/math11061523 | |
8. | A. M. Elaiw, E. A. Almohaimeed, A. D. Hobiny, Modeling the co-infection of HTLV-2 and HIV-1 in vivo, 2024, 32, 2688-1594, 6032, 10.3934/era.2024280 | |
9. | Siyu Chen, Zhijun Liu, Lianwen Wang, Xingan Zhang, Global dynamics analysis for a nonlinear HTLV-I model with logistic proliferation and CTL response, 2024, 17, 1793-5245, 10.1142/S1793524523500237 | |
10. | Sourav Chowdhury, Jayanta Kumar Ghosh, Uttam Ghosh, Co-infection dynamics between HIV-HTLV-I disease with the effects of Cytotoxic T-lymphocytes, saturated incidence rate and study of optimal control, 2024, 223, 03784754, 195, 10.1016/j.matcom.2024.04.015 | |
11. | Rasha Majeed Yaseen, Hassan Fadhil Al-Husseiny, 2024, 3061, 0094-243X, 040040, 10.1063/5.0196252 | |
12. | Zeynab Hemmatzadeh, Vahid Roomi, Tohid Kasbi Gharahasanlou, Stability, Hopf Bifurcation and Numerical Simulation of an HIV Model with Two Modes of Transmission and with Cellular and Humoral Immunity, 2023, 33, 0218-1274, 10.1142/S0218127423501626 | |
13. | A.M. Elaiw, E.A. Almohaimeed, A.D. Hobiny, Stability analysis of a diffusive HTLV-2 and HIV-1 co-infection model, 2025, 116, 11100168, 232, 10.1016/j.aej.2024.11.074 | |
14. | Ayesha Saleem, Mati ur Rahman, Salah Boulaaras, Rafik Guefaifia, Dumitru Baleanu, Exploring the dynamics of HIV and HCV co-infection through piecewise modified Mittag-Leffler fractional derivatives, 2025, 33, 2769-0911, 10.1080/27690911.2025.2478038 | |
15. | E. A. Almohaimeed, A. M. Elaiw, A. D. Hobiny, Modeling HTLV-1 and HTLV-2 co-infection dynamics, 2025, 10, 2473-6988, 5696, 10.3934/math.2025263 | |
16. | A. M. Elaiw, E. A. Almohaimeed, Within-host dynamics of HTLV-2 and HIV-1 co-infection with delay, 2025, 19, 1751-3758, 10.1080/17513758.2025.2506536 | |
17. | Sourav Chowdhury, Biswajit Paul, Uttam Ghosh, Effect of Stochastic Transition in HIV Infection Model With Cytotoxic T‐Lymphocytes and Beddington–Deangelis Incidence and Linear Cure Rate, 2025, 0170-4214, 10.1002/mma.11114 |