Hindawi

International Journal of Differential Equations

Volume 2020, Article ID 5051248, 14 pages
https://doi.org/10.1155/2020/5051248

Research Article
Stability and Global Sensitivity Analysis for an Agree-Disagree
Model: Partial Rank Correlation Coefficient and Latin Hypercube
Sampling Methods

Sara Bidah , Omar Zakary , and Mostafa Rachik

1
Laboratory of Analysis Modelling and Simulation, Department of Mathematics and Informatics, Faculty of Sciences Ben M’Sik,
Hassan II University of Casablanca, BP 7955, Sidi Othman, Casablanca, Morocco

Correspondence should be addressed to Omar Zakary; [email protected]

Received 1 January 2020; Accepted 2 March 2020; Published 1 April 2020

Academic Editor: Mayer Humi

Copyright © 2020 Sara Bidah et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

In this paper, we present a new mathematical model that describes agree-disagree opinions during polls. We first present the
model and its different compartments. Then, we use the next-generation matrix method to compute thresholds of equilibrium
stability. We perform the stability analysis of equilibria to determine under which conditions these equilibrium points are stable or
unstable. We show that the existence and stability of these equilibria are controlled by the calculated thresholds. Finally, we also
perform several computational and statistical experiments to validate the theoretical results obtained in this work. To study the
influence of various parameters on these thresholds and to identify the most influential parameters, a global sensitivity analysis is
carried out based on the partial rank correlation coefficient method and the Latin hypercube sampling.

1. Introduction Polls then play a key role in contemporary political

campaigns. Party performance updates receive considerable
Participation in political life requires citizens’ attention, media attention and often serve as a basis for political
time, knowledge, money, and motivation. Citizens will commentary in the weeks leading up to the election day. We
participate if they receive benefits commensurate with the know that knowing the positions of the electorate can affect
cost of participation. Perhaps, the debate on voting is the voter behavior [3, 4]. However, it is unclear how voting
subject of the most studied political behavior. Individual shapes party leaders’ strategies during the election campaign.
economic, social, and psychological costs and voting benefits Public opinion is often criticized for influencing voter
are well known [1]. Candidates and the media often use polls perceptions and focusing on popularity rather than politics.
in election campaigns to determine which candidates are in By strategically communicating the poll results, parties can
the lead and who are likely to emerge victoriously. Opinion shape voters’ vision of competition and encourage popular
polls are surveys of intent for a sample of voters, while future mobilization [5]. This explains why, at the beginning of the
results allow participants to negotiate and discuss opinions 20th century, more than thirty democratic countries banned
based on a particular outcome. Opinion polls revealed the the publication of opinion polls near elections [6, 7].
expected voting quotas. Public opinion polls are widely used To fully understand the role of polls before the elections,
to identify people’s political positions, vote, and other be- it is necessary to take into account the offer of campaigns.
haviors by asking questions about their opinions, activities, This is particularly important in contexts of increased po-
and personal characteristics. Answers to these questions are larization, where holidays play an increasing role in deter-
then counted, statistically analyzed, and interpreted. Polls’ mining the type of information that reaches voters [8]. At
result also provides parties with the opportunity to refine any time, party leaders may choose to speak of candidates
their campaign strategies. In the long run, parties change from their party or other candidates for election, focusing on
their positions in public opinion [2]. various political or gender issues. Voter signals in opinion
2 International Journal of Differential Equations

polls can influence party leaders’ decisions to balance these example, if there are four parts A, B, C, and D, we wish to
different elements of campaign speech [9]. study the political position of party A, and we can then
In the electoral process, where there are only two po- examine the two subsets {A} and {B, C, D} of the investi-
litical parties, the chances of creating strong political po- gation. Votes for A are considered agree, and votes for B, C,
larization are increasing. For example, U. S. citizens face a or D are considered disagree. More simply, we consider the
polarizing scenario during a presidential election, where opinion poll of voters on the performance of the party
they must vote for a two-party decision: Republicans and studied. As an example, we can cite the “approve or dis-
Democrats. Previous studies have shown that U. S. elections approve” poll done by Thomson Reuters on how Donald
are attracting the Internet, where both Twitter and blogs are Trump fulfills his role as the president [15].
showing high political polarization [10, 11]. Without loss of generality, we devise a mathematical
Other political scenarios requiring a bilateral decision model describing the evolution of agree and disagree
are second-round electoral processes. While citizens of opinions during polls (surveys), and the types of surveys we
the first round can vote from a wide range of different consider here are surveys that can be answered in agreement,
political parties, in the last round, they can vote only for disagreement, or otherwise. Thus, the population targeted by
the two final candidates. Voters may not be fully identified the poll is regrouped into three groups: agree, disagree, and
with either side, but they still have to take a side. Previous ignorant individuals.
articles have shown that this second round increases The model herewith has been formulated using three
political polarization in the country [12]. In [13], the compartments. Each of them has been described as follows:
authors analyzed the 2017 presidential election in Chile
and measured the resulting political polarization. Of a (1) Ignorant (I): people who do not know about the poll
minority of qualified users, they could estimate the or those who abstain from voting for personal
opinion of the majority. reasons
In this contribution, we examine how voters’ preferences (2) Agree (A): people in agreement with the idea being
and expectations compete with each other when dealing with studied
voter support information to candidates. So, we start by (3) Disagree (D): people in disagreement with the idea
developing a mathematical model to describe the evolution being studied
of opinions to predict the probability of results, and then we
calculate and analyze the equilibrium points of the model by All contacts are modeled by the standard incidence rate.
deriving the important stability thresholds for each equi- For the modeling processes, a set of assumptions has been
librium state. used. These are as follows:
Through the development of more efficient models, (1) The targeted population is well mixed, that is, the
reliable statistical and mathematical methods are needed to ignorant individuals are homogeneously spread
improve the accuracy of modeling. One of the most recently throughout the population
popular ways is sensitivity analysis (SA) methods. Sensitivity
analysis is used for a variety of reasons, such as developing (2) Recruitment and mortality are negligible under the
decisions or recommendations, communicating, under- temporal scale consideration; therefore, no indi-
standing or quantifying the system, and developing models. vidual is recruited and no individual dies during the
In the development of the model, it can be used to verify the poll
validity or accuracy of the model, to simplify, calibrate, weak (3) Individuals have the right to communicate with each
process, or lost data, and even to identify the important other and can thus convince one another
parameter for other studies [14]. (4) People who are unsure of their opinion are ignorant
The paper is organized as follows: Section 2 introduces
(5) People who abstain from voting are ignorant
our new model, giving some details about interactions be-
tween different compartments and parameters of the model. Everyone has their reasons for agreement or dis-
In Section 3, we derive basic reproduction numbers. Section agreement. An ignorant person can be persuaded by
4 provides the results of the stability analysis of equilibria. In someone who agrees at a rate β1 or by someone who does
Section 5, sensitivity analysis is performed to identify the not agree with the opinion at a rate β2. A person agreeing
most important parameter in the proposed model, and with the opinion may be persuaded by someone who does
Section 6 concludes the paper. not agree at a rate α2, or a person who disagrees with the
opinion may be persuaded by someone who agrees at a
2. Presentation of the Model rate α1. People can abstain from voting or lose interest
without any direct contact with individuals from the
There are many scenarios involving a binary decision. The opposite opinion group, then agree people become ig-
poll model that we present here describes the opinions of norants at the rate c1, and disagree people become ig-
agreement (or approval) and disagreement (or disapproval), norants at the rate c2. A flowchart describing different
regarding a candidate or idea, in polls preceding the elec- interactions between the compartments of the model is
tions. Note that the model can also describe situations in presented in Figure 1.
which there are more than two candidate parties because we All these assumptions and considerations are written as
can always reduce the situation to two decisions. For the following system of ordinary differential equations:
International Journal of Differential Equations 3

AI DI β2(DI/N)
I′ � − β1 − β2 + c1 A + c2 D, (1)
N N I D
γ2D
AI AD AD
A ′ � β1 + α1 − α2 − c1 A, (2)
N N N β1(AI/N) α1(AD/N)
DI AD AD γ1A α2(AD/N)
D′ � β2 + α2 − α1 − c2 D, (3)
N N N
A
where I(0) ≥ 0, A(0) ≥ 0, D(0) ≥ 0, and N � I + A + D. Note
that N′ � I′ + A′ + D′ � 0; thus, the population size N is
Figure 1: Flowchart for models (1)–(3).
considered as a constant in time. We can easily prove that,
for nonnegative initial conditions, the solutions of systems
(1)–(3) are nonnegative. To do this, recall that by [16] the
system of equation Table 1: Parameter descriptions.
x′ � f x1 , x2 , . . . , xk 􏼁, (4) Parameter Description
β1 Ignorant to agree transmission rate
with β2 Ignorant to disagree transmission rate
x(0) � x0 ≥ 0, (5) α1 Disagree to agree transmission rate
α2 Agree to disagree transmission rate
is a positive system if and only if ∀i � 1, 2, . . ., k. c1 Interest loss factor of agree individuals
c2 Interest loss factor of disagree individuals
x′i � fi x1 ≥ 0, x2 ≥ 0, . . . , xi � 0, . . . , xk ≥ 0􏼁 ≥ 0. (6)

Thus, for models (1)–(3), it is easy to verify that

growth factor [22]. Generally, if R0 > 1, an epidemic
I � 0 ⟹ I′ ≥ 0,
occurs, whereas if R0 < 1, there will probably be no
A � 0⟹ A′ ≥ 0, (7) epidemic.
D � 0 ⟹ D′ ≥ 0. Following this definition, we will define our thresholds
as follows: RD0 is the average number of new disagree-
Therefore, all the solutions of systems (1)–(3) are ments produced by an individual disagreeing introduced
nonnegative. in a population of ignorant people during the period in
It is also clear that the solutions of models (1)–(3) are which he or she was in this opinion. And, RA0 is the
bounded based on the fact that N � I + A + D is constant, and average number of new agreements produced by an in-
then S ≤ N, A ≤ N, and D ≤ N. Therefore, we will focus to dividual in agreement and that was introduced into a
study models (1)–(3) in the closed positively invariant population of ignorant people during the period in which
feasible set given by he or she was in that opinion.
For the analysis of epidemic models, the first step is to
(I, A, D) ∈ R3+
Ω �􏼠 􏼡. (8) calculate the disease-free equilibrium (DFE). This equilib-
I+A+D�N rium point is then used to calculate the basic reproductive
number using the next-generation matrix method. The
A summary of parameter description is given in Table 1.
objective of this section is only the calculation of the
thresholds and not the equilibrium states of the model. But,
3. Thresholds: Basic Reproductive Numbers in this method, we have to determine the equilibrium states
when A � 0 and when D � 0.
In epidemiology, the basic reproductive number R0 (or In this contribution, let RX0 be the threshold of growing
epidemic threshold) is defined as the average number of of the opinion X (either X � A “agree” or X � D “disagree”).
secondary cases of an infection produced by a “typical” Then, RD0 is the threshold associated with the disagree-free
infected individual during his/her entire life as infectious equilibrium, while RA0 is the one associated with the agree-
when introduced in a population of susceptibles [17–21]. free equilibrium.
The threshold R0 is mathematically characterized in terms We calculate the equilibria for the aforementioned
of infection transmission as a “demographic process,” but model, and based on the next-generation matrix approach,
offspring production is not seen as giving birth in a de- we derive associated thresholds.
mographic sense, but it causes new infections through The points of equilibrium of systems (1)–(3) are the
transmission [22–24]. Thus, the infection process can be solutions of
considered as a “consecutive generation of infected in-
dividuals.” The following growing generations indicate a I′ � A′ � D′ � 0, (9)
growing population (i.e., an epidemic), and the growth
factor for each generation indicates the potential for for the disagree-free equilibrium when there is no negative
growth. So, the mathematical characterization of R0 is this opinion, i.e., if we put D � 0. This gives
4 International Journal of Differential Equations

c1 α2 β1 − α2 c1 + β2 c1
I∗ � N, RD0 � > 0,
β1 α1 β1 − α1 c1 + β1 c2
(10) (17)
N β1 − c1 􏼁
∗ α β − α1 c 2 + β 1 c 2
A �
β1
, RA0 � 1 2 > 0.
α2 β2 − α2 c2 + β2 c1
where I∗ and A∗ represent the numbers of ignorant and
agree individuals, respectively, in the absence of disagree
people. 4. Stability Analysis
Therefore, for the system governed by (1)–(3), the dis- In order to analyze in terms of the proportions of ignorant,
agree-free equilibrium is agree, and disagree individuals, let i � (I/N), a � (A/N),
c N β1 − c1 􏼁 and d � (D/N) denote the fraction of the classes I, A, and D
e1 � 􏼠 1 N, , 0􏼡. (11)
β1 β1 in the population, respectively. After some calculations and
replacing I by i, A by a, and D by d, equations (1)–(3) can be
Following the second generation approach [22], we written as
compute the threshold RD0 associated to the disagree-free
i′ � − β1 ai − β2 di + c1 a + c2 d, (18)
equilibrium, which is
α β − α2 c1 + β2 c1 a′ � β1 ai + α1 ad − α2 ad − c1 a,
RD0 � 2 1 (12) (19)
α1 β1 − α1 c1 + β1 c2
for the agree-free equilibrium when there is no positive d′ � β2 di + α2 ad − α1 ad − c2 d. (20)
opinion, i.e., if we put A � 0. This gives
From the fact N � I + A + D, we have i + a + d � 1. Then,
c
I∗ � 2 N, model systems (18)–(20) will be reduced to the following two
β2 differential equations:
(13)
N β2 − c 2 􏼁 a′ � β1 a(1 − a − d) + α1 ad − α2 ad − c1 a,
∗ (21)
D � ,
β2 d′ � β2 d(1 − a − d) + α2 ad − α1 ad − c2 d,
where I∗ and D∗ represent the numbers of ignorant and which can be reduced to
disagree individuals, respectively, in the absence of agree
a′ � β1 a(1 − a − d) + αa c1 a,
people. (22)
Therefore, for the system governed by (1)–(3), the agree- d′ � β2 d(1 − a − d) − αad − c2 d,
free equilibrium is
where α � α1 − α2.
c2 N β2 − c2 􏼁
e2 � 􏼠 N, 0, 􏼡. (14)
β2 β2
4.1. Steady States. The steady states of system (22) are ob-
By also following the second generation approach [22], tained by solving the system of equations
we compute the threshold RA0 associated to the agree-free 0 � β1 a(1 − a − d) + αad − c1 a,
equilibrium, which is (23)
0 � β2 d(1 − a − d) − αad − c2 d.
α1 β 2 − α 1 c 2 + β 1 c 2
RA0 � . (15) This system has four equilibrium points, and the trivial
α2 β 2 − α 2 c 2 + β 2 c 1
equilibrium E0 � (0, 0) is an equilibrium that exists always
Throughout this paper, we consider the following as- without any condition. It means there is no survey and there
sumption: the model parameters verify is no need to opinions.
One disagree-free equilibrium E1 � (1 − (c1 /β1 ), 0) ex-
βc
β 1 − c1 > − 1 2 , ists if the condition β1 > c1 holds, and the agree-free
α1 equilibrium E2 � (0, 1 − (c2 /β2 )) exists if β2 > c2 holds.
The fourth and the positive equilibrium E∗ � (a∗ , d∗ ),
β 2 c1
β1 − c1 > − , where
α2
(16) αβ − αc2 + β1 c2 − β2 c1
a∗ � 2 ,
βc α α − β1 + β2 􏼁
β2 − c2 > − 2 1 , (24)
α2
∗ αβ1 − αc1 + β1 c2 − β2 c1
d �− ,
β 1 c2 α α − β 1 + β2 􏼁
β2 − c2 > − .
α1
exists if one of the following conditions holds:
Therefore, from the above assumption, we get “RD0 > 1 and RA0 > 1 and α(α − β1 + β2 ) > 0” or “RD0 < 1
International Journal of Differential Equations 5

and RA0 < 1 and α(α − β1 + β2 ) < 0.” In fact, by a simple It is clear from the existence condition of this equilib-
calculation, we get rium that c1 − β1 < 0. Thus, the stability of the point of
αβ2 − αc2 + β1 c2 − β2 c1 > 0⟺RA0 > 1, equilibrium E1 is based on the eigenvalue λ2 of the matrix
(25) J(E1 ). By a simple calculation, we have
αβ1 − αc1 + β1 c2 − β2 c1 < 0⟺RD0 > 1.
αβ1 − αc1 + β1 c2 − β2 c1 < 0⟺RD0 > 1. (31)
∗ ∗
Thus, from (25), we deduce that a > 0 and d > 0.
This implies that if RD0 > 1, E1 is unstable; else, E1 is
Throughout the article and without loss of generality, we just
consider the first condition as a sufficient condition for the
stable. □
existence of E∗ .
Proposition 3. The equilibrium E2 � (0, 1 − (c2 /β2 )) is
unstable if RA0 > 1. Otherwise, it is stable.
4.2. Stability of Steady States. The Jacobian matrix of system
(22) is Proof. The Jacobian matrix at this equilibrium is
J11 aα − aβ1 β 1 c2 c
J �􏼠 􏼡, (26) ⎜
⎛
⎜ − α􏼠 2 − 1 􏼡 − c 1 0 ⎞ ⎟
⎟
− αd − β2 d J22 ⎜
⎜
⎜ β2 β2 ⎟
⎟
⎟
⎜
⎜
⎜ ⎟
⎟
⎟
J E2 􏼁 � ⎜
⎜
⎜ ⎟
⎟
⎟ . (32)
where ⎜
⎜
⎜ ⎟
⎟
⎟
⎜
⎝ α + β 2 􏼁 β2 − c2 􏼁 ⎟
⎟
⎠
J11 � αd − aβ1 − c1 − β1 (a + d − 1), − c2 − β 2
(27) β2
J22 � − c2 − aα − β2 d − β2 (a + d − 1).
The eigenvalues of J(E2 ) are
β 1 c2 c
Proposition 1. The equilibrium E0 � (0, 0) is unstable if λ1 � − α􏼠 2 − 1􏼡 − c1
β1 > c1 or β2 > c2. Otherwise, it is stable. β2 β2

αβ2 − αc2 + β1 c2 − β2 c1 (33)

Proof. The Jacobian matrix at this equilibrium is � ,
β2
β 1 − c1 0
J E0 􏼁 � 􏼠 􏼡. (28)
0 β2 − c 2 λ 2 � c2 − β2 .

It is clear that if β1 > c1 or β2 > c2, we get one positive It is clear from the existence condition of this equilib-
eigenvalue of J(E0 ), and then E0 is unstable. Else, we have all rium that c2 − β2 < 0. Thus, the stability of the point of
eigenvalues of J(E0 ) having the negative real part, which equilibrium E2 is based on the eigenvalue λ1 of the matrix
completes the proof. □ J(E2 ). By a simple calculation, we have
αβ2 − αc2 + β1 c2 − β2 c1 > 0⟺RA0 > 1. (34)
Remark 1. Note that the conditions in the previous prop-
osition imply the existence of E1 or E2. Therefore, E0 is This implies that if RA0 > 1, E2 is unstable; else, E2 is
unstable whenever there exists E1 or E2. stable. □
Proposition 2. The equilibrium E1 � (1 − (c1 /β1 ), 0) is Proposition 4. The equilibrium E∗ � (a∗ , d∗ ), where
unstable if RD0 > 1. Otherwise, it is stable. αβ − αc2 + β1 c2 − β2 c1
a∗ � 2 ,
Proof. The Jacobian matrix at this equilibrium is α α − β1 + β2 􏼁
(35)
α − β1 􏼁 β1 − c1 􏼁 ∗ αβ − αc1 + β1 c2 − β2 c1
⎜
⎛
⎜ c1 − β1 ⎟
⎞
⎟ d �− 1 ,
⎜
⎜
⎜ β1 ⎟
⎟
⎟ α α − β 1 + β2 􏼁
⎜
⎜ ⎟
⎟
⎟
⎜
⎜
J E1 􏼁 � ⎜ ⎟
⎟ . (29)
⎜
⎜ ⎟
⎟
⎟ is stable if “β1 < c1 and c2 < β2 and α > 0” or “β1 > c1 and c2 >
⎜
⎜ ⎟
⎜
⎝ c1 β2 c1 ⎟
⎟
⎠ β2 and α < 0.”
0 α􏼠 − 1􏼡 − c2 +
β1 β1
Proof. The Jacobian matrix at this equilibrium is
The eigenvalues of J(E1 ) are
J11 a∗ α − a∗ β1
λ1 � c1 − β1 , J E∗ 􏼁 � 􏼠 􏼡, (36)
− αd∗ − β2 d∗ J22
c1 βc where
λ2 � α􏼠 − 1􏼡 − c2 + 2 1
β1 β1 (30) J11 � αd∗ − a∗ β1 − c1 − β1 a∗ + d∗ − 1􏼁,
(37)
αβ − αc1 + β1 c2 − β2 c1 J22 � − c2 − a∗ α − β2 d∗ − β2 (a + d − 1).
�− 1 .
β1 Using fact (23), we have
6 International Journal of Differential Equations

J11 � − β1 a∗ , (2) Sufficient conditions in the above proposition could

∗ (38) be reduced to α (β1c2 – β1c2) < 0.
J22 � − β2 d . A summary of sufficient existence and stability condi-
tions of the steady states of model (11) is given in Table 2.
Then,
J1 J2
J E∗ 􏼁 � ⎛
⎝ ⎠,
⎞
4.3. Examples. It can be seen from Table 3 that the equi-
J3 J4 librium E0 exists and it is stable in the four cases: example 1:
β1 αβ2 − αc2 + β1 c2 − β2 c1 􏼁 RD0 > 1 and RA0 > 1, example 2: RA0 < 1 and RD0 < 1, example
J1 � − , 3: RD0 < 1 and RA0 > 1, and example 4: RD0 > 1 and RA0 < 1.
α α − β1 + β2 􏼁
Example 5 shows the existence and the stability of the
α − β1 􏼁 αβ2 − αc2 + β1 c2 − β2 c1 􏼁 (39) steady state E1 in the case of RA0 < 1 and RD0 < 1 and in the case
J2 � , of RA0 > 1 and RD0 < 1 in example 6. Examples 7 and 8 show the
α α − β1 + β2 􏼁
existence and the stability of the equilibrium E2 in the cases of
α + β2 􏼁 αβ1 − αc1 + β1 c2 − β2 c1 􏼁 “RA0 < 1 and RD0 < 1” and “RA0 < 1 and RD0 > 1,” respectively.
J3 � ,
α α − β1 + β2 􏼁 Examples 9 and 10 show the possibility of the existence and
stability of the two steady states E1 and E2 at the same time, that
β2 αβ1 − αc1 + β1 c2 − β2 c1 􏼁 is, “β1 > c1 and β2 > c2 and RA0 < 1 and RD0 < 1.” These
J4 � ,
α α − β1 + β2 􏼁 examples give an insight about the stability of each equilibrium
for given parameters’ values. After some numerical calcula-
where the characteristic polynomial of is tions, we noted that, in this situation, the parameters c1 and c2
λ2 + c1 λ + c2 , (40) may switch from E1 to E2 and the inverse, while if c1 > c2, then
E1 will be more attractive, and if c2 > c1, then E2 will be more
where attractive and that with the same initial conditions. We sim-
β 1 c2 − β 2 c1 ulate the model with different initial conditions to illustrate the
c1 � − , impact of initial conditions on the stability of E1 and E2 in this
α
situation. Figure 2 depicts the stability of E1 and E2 at the same
−1 time, where we consider the same set of parameters from
c2 �
α α − β1 + β2 􏼁 example 9 in Table 3. By changing the initial conditions, we can
see that E1 is stable when I(0) � 100, A(0) � 100, and
× 􏼐α2 β1 β2 − α2 β1 c2 − α2 β2 c1 + α2 c1 c2 D(0) � 80, while it can be seen that E2 is also stable with this set
of parameters but after choosing I(0) � 100, A(0) � 80, and
+ αβ21 c2 − αβ1 β2 c1 + αβ1 β2 c2 − αβ1 c1 c2 D(0) � 100.
Example 11 shows the existence and the stability of the
− αβ1 c22 − αβ22 c1 + αβ2 c21 + αβ2 c1 c2 equilibrium E∗ , where “RD0 > 1 and RA0 > 1 and α(α − β1 +
β2 ) > 0” and “β1 < c1 and c2 < β2 and α > 0,” while example 12
+ β21 c22 − 2β1 β2 c1 c2 + β22 c21 􏼑, shows the existence and the stability of E∗ , where
“RD0 > 1 and RA0 > 1 and α(α − β1 + β2 ) > 0” and “β1 > c1
−1 and c2 > β2 and α < 0.”
�
α α − β1 + β2 􏼁 Figure 3 depicts examples of the existence and stability of
the equilibrium state E0 for different parameters’ values and
× αβ1 − αc1 + β1 c2 − β2 c1 􏼁 αβ2 − αc2 + β1 c2 − β2 c1 􏼁. threshold RD0 and RA0 values simulated with initial con-
(41) ditions and parameters’ values from Table 3. Figure 4 depicts
examples of the existence and stability of the equilibrium
By the conditions “β1 < c1 and c2 < β2 and α > 0” or “β1 > state E1 for different parameters’ values and threshold RD0
c1 and c2 > β2 and α < 0,” we have − (β1 c2 − β2 c1 /α) > 0, and and RA0 values simulated with initial conditions and pa-
by existence conditions and rameters’ values from Table 3. It can be seen from
αβ2 − αc2 + β1 c2 − β2 c1 > 0⟺RA0 > 1, Figure 4(b) that the function D decreases towards zero, but it
(42) will take a long time. In the Figure 4(c), we have considered
αβ1 − αc1 + β1 c2 − β2 c1 < 0⟺RD0 > 1, 1,200 hours to show that the function D will go to zero but
we have c1 > 0 and c2 > 0. Using the Routh–Hurwitz stability very slowly. Figure 5 depicts examples of the existence and
criterion, we conclude that the equilibrium point E∗ is lo- stability of the equilibrium state E2 for different parameters’
cally asymptotically stable. □ values and threshold RD0 and RA0 values simulated with
initial conditions and parameters’ values from Table 3. We
can see from Figure 5(b) that the function A will also take a
Remark 2 very long time to go to zero. In Figure 5(c), we have con-
sidered 1,200 hours to show that function A will tend to zero
(1) Note that the conditions RD0 > 1 and RA0 > 1 lead to the but very slowly. Figure 6 depicts examples of the existence
instability of E1 and E2 and help in the existence of E∗ . and stability of the equilibrium state E∗ for different
International Journal of Differential Equations 7

Table 2: Summary of sufficient existence and stability conditions of equilibria.

Equilibrium Existence conditions Stability conditions
E0 — β1 < c1 and β2 < c2
E1 � (1 − (c1 /β1 ), 0) β1 > c1 RD0 < 1
E2 � (0, 1 − (c2 /β2 )) β2 > c2 RA0 < 1
E∗ � (a∗ , d∗ ) RD0 > 1 and RA0 > 1 and α(α − β1 + β2 ) > 0 β1 < c1 and c2 < β2 and α > 0 or β1 > c1 and c2 > β2 and α < 0

Table 3: Steady states and parameters’ values used in example simulations, where I(0) � A(0) � D(0) � 100.
β1 β2 α1 α2 c1 c2 RD0 RA0
1 E0 � (300, 0, 0) 0.0010 0.1010 0.1010 0.3010 0.5010 0.3010 1.9901 2.0730
2 E0 � (300, 0, 0) 0.0010 0.2010 0.2010 0.3010 0.1010 0.5010 0.5000 0.8543
3 E0 � (300, 0, 0) 0.0010 0.0010 0.2010 0.1010 0.1010 0.1010 0.5000 2.0001
4 E0 � (300, 0, 0) 0.0010 0.0010 0.1010 0.2010 0.1010 0.1010 2.0001 0.5000
5 E1 � (2.9703, 297.0297, 0) 0.1010 0.0010 0.2010 0.2010 0.0010 0.0010 0.6634 0.4925
6 E1 � (2.9703, 297.0297, 0) 0.1010 0.0010 0.0010 0.0010 0.0010 0.0010 0.5025 101
7 E2 � (2.9703, 0, 297.0297) 0.0010 0.1010 0.2010 0.2010 0.1010 0.0010 0.4925 0.6634
8 E2 � (2.9703, 0, 297.0297) 0.0010 0.1010 0.0010 0.0010 0.0010 0.0010 101 0.5025
9 E∗1 � (120.3593, 179.6407, 0) 0.5010 0.1010 0.1010 0.0010 0.2010 0.0010 0.6688 0.5196
10 E∗2 � (120.3593, 0, 179.6407) 0.1010 0.5010 0.0010 0.1010 0.0010 0.2010 0.5196 0.6688
11 E∗ � (133.3333, 10.5556, 156.1111) 0.1010 0.7010 0.3010 0.0010 0.2010 0.3010 467.7774 1.0672
12 E∗ � (214.2857, 80.4643, 5.2500) 0.3010 0.4010 0.0010 0.8010 0.2010 0.5010 1.0649 300.8004

300 300
Number of people
Number of people

250 250
E1 = (112, 168, 0) E2 = (2, 0, 278)
200 200
150 150
100 100
50 50
0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Time (hours) Time (hours)
I I
A A
D D
(a) (b)

Figure 2: Illustration of the bistability of E1 and E2 at the same time using the set of parameters of example 9 from Table 3. Stability of E1 (a)
when I(0) � 100, A(0) � 100, and D(0) � 80. Stability of E2 (b) when I(0) � 100, A(0) � 80, and D(0) � 100.

parameters’ values and threshold RD0 and RA0 values sim- E1 � (1, 0),
ulated with initial conditions and parameters’ values from (44)
examples 11 and 12 of Table 3. E2 � (0, 1),

and from the existence conditions in Table 2, we can deduce

5. Analysis of Thresholds that there is no E∗ .
E1 exists without any condition, and it is stable if α2 < α1,
5.1. No Abstention: No Loss of Interest. In most of the sit- and E2 exists without any condition, and it is stable if α1 < α2.
uations, people abstain from voting for personal reasons. This result explains the effect of the polarization on the
Here, we discuss the situation when there is no abstention outcome of polls. When there is no abstention from voting
and there is no loss of interest of voting, i.e., c1 � c2 � 0. and people keep their interest in voting, then the most
Therefore, RD0 and RA0 become influential parameters on the outcome of the polls are the
α2 polarization factors α1 and α2. For example, during a poll
RD0 � , determining the political position of one candidate Y by
α1
(43) voting with Approve or Disapprove or otherwise, if can-
α didate Y can convince people by his political vision and/or
RA0 � 1, by other ways, then he can change the course of events in his
α2
favor. Mathematically, he makes α2 < α1. But, if somehow he
which means that thresholds RD0 and RA0 depend only on α1 contributes to making α1 < α2, then things could spin out of
and α2, and the equilibria become control on the election day.
8 International Journal of Differential Equations

300 300
250 250
Number of people

Number of people
200 200
150 150
100 100
50 50
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (hours) Time (hours)
I I
A A
D D
(a) (b)
300 300
250 250

Number of people
Number of people

200 200
150 150
100 100
50 50
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (hours) Time (hours)
I I
A A
D D
(c) (d)

Figure 3: Examples of the steady state E0 simulated with initial conditions I(0) � 100, A(0) � 100, and D(0) � 100 and parameters’ values
from Table 3. (a) Example 1. (b) Example 2. (c) Example 3. (d) Example 4.

300 300
250 250
Number of people

Number of people

200 200
150 150
100 100
50 50
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (hours) Time (hours)
I I
A A
D D
(a) (b)
Figure 4: Continued.
International Journal of Differential Equations 9

300
250

Number of people
200
150
100
50
0
0 200 400 600 800 1000
Time (hours)
I
A
D
(c)

Figure 4: Examples of the steady state E1 simulated with initial conditions I(0) � 100, A(0) � 100, and D(0) � 100 and parameters’ values
from Table 3. (a) Example 5. (b) Example 6. (c) Example 9.

300 300
250 250
Number of people

Number of people
200 200
150 150
100 100
50 50
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (hours) Time (hours)
I I
A A
D D
(a) (b)
300
250
Number of people

200
150
100
50
0
0 200 400 600 800 1000
Time (hours)
I
A
D
(c)

Figure 5: Examples of the steady state E2 simulated with initial conditions I(0) � 100, A(0) � 100, and D(0) � 100 and parameters’ values
from Table 3. (a) Example 7. (b) Example 8. (c) Example 10.

5.2. One Chance. In some situations, voters are not allowed β 2 c1

to modify their choice; then, they could take one side until RD0 � ,
β 1 c2
the end of the survey. In this section, we discuss the sit- (45)
uation when there is no persuasion and then there is no βc
polarization, i.e., α1 � α2 � 0. Therefore, RD0 and RA0 RA0 � 1 2,
β 2 c1
become
10 International Journal of Differential Equations

300 300
250 250
Number of people

Number of people
200 200
150 150
100 100
50 50
0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time (hours) Time (hours)

I I
A A
D D
(a) (b)

Figure 6: Examples of the steady state E∗ simulated with initial conditions I(0) � 100, A(0) � 100, and D(0) � 100 and parameters’ values
from Table 3. (a) Example 11. (b) Example 12.

which means that thresholds RD0 and RA0 depend only on β1, Table 4: Six input parameters’ probability distribution functions.
β2, c1, and c2, and there is no change in the equilibria E1 and
Parameters Pdf
E2. From the conditions of the existence in Table 2, we can
deduce that there is no E∗ . A sufficient condition to make E1 β1 Normal
β2 Normal
stable is “β2 < β1 and c1 < c2,” and the sufficient condition to
α1 Triangular
make E2 stable is “β2 > β1 and c1 > c2.” This result explains α2 Triangular
the efficiency of election campaigns. For instance, if there are c1 Triangular
two candidate parts X and Y and the poll is carried out to c2 Triangular
study the political position of Y, then votes for Y are con-
sidered as agree and votes for X are considered as disagree. If
candidate Y presents a successful election campaign, it will
densely stratifies the input parameters. As the name suggests,
attract more people alongside him and increase the number
the PRCC measures the strength between the inputs and
of people agreeing (i.e., β2 < β1), which may lead disagree
outputs of the model using correlation through the sampling
people to lose interest or abstain from voting (i.e., c1 < c2).
done by the LHS method [26–28].
Parameters β1 and β2 follow normal distribution with
5.3. Statistical Analysis. Here, we use probability distribu- mean and standard deviation 0.5 and 0.01, respectively, while
tion functions of the six parameters given in Table 4 sampled parameters α1, α2, c1, and c2 follow triangular distribution
by using the Latin hypercube sampling, see Figure 7. We with minimum, maximum, and mode as 0.02, 0.8, and 0.51,
compute the probabilities of equilibria existence and stability respectively. A summary of probability distribution functions
conditions. It can be seen from Table 5 that the probability of is given in Table 4. In Latin hypercube sampling approach,
E1 to exist is about 0.5960, while its probability of stability is probability density function (given in Table 4) for each pa-
about 0.53. The probability of E1 to exist and to be stable is rameter is stratified into 100 equiprobable (1/100) serial in-
0.4040. The probability of the existence of E2 is about 0.6250, tervals. Then, a single value is chosen randomly from each
while the probability of its stability is about 0.5370. The interval. This produces 100 sets of values of RD0 and RA0, from
probability of E1 to exist and to be stable is 0.43. The 100 sets of different parameter values mixed randomly, cal-
equilibrium E∗ has a probability of existence about 0.0240 culated by using equations (4) and (5), respectively.
and a probability of stability about 0.2190, while E∗ exists, Sensitivity analysis has been done concerning the basic
and it is stable with a probability of 0.001. reproduction numbers RD0 and RA0, and the main objective
of this section is to identify which parameters are important
6. Sensitivity Analysis in contributing variability to the outcome of basic repro-
duction numbers based on their estimation uncertainty.
Global sensitivity analysis (GSA) approach helps to identify To sort the model parameters according to the size of
the effectiveness of model parameters or inputs and thus their effect on RD0 and RA0, a partial rank correlation
provides essential information about the model perfor- coefficient is calculated between the values of each of the six
mance. Out of the many methods of carrying out sensitivity parameters and the values of RD0 and RA0 in order to
analysis is the partial rank correlation coefficient (PRCC) identify and measure the statistical influence of any one of
method that has been used in this paper. The PRCC is a the six input parameters on thresholds RD0 and RA0. The
method based on sampling. One of the most efficient larger the partial rank correlation coefficient, the larger the
methods used for sampling is the Latin hypercube sampling influence is on the input parameter affecting the magnitude
(LHS), which is a type of Monte Carlo sampling [25] as it of RD0 and RA0.
International Journal of Differential Equations 11

0.8 1 8
0.7
0.8
6
0.6
0.6
0.5
4
0.4 0.4

0.3 2
0.2
0.2
0 200 400 600 800 1000 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
(a) (b) (c)
8 8 8

6 6 6

4 4 4

2 2 2

0 0 0
0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
(d) (e) (f )

Figure 7: Probability distribution functions of the six sampled input parameters’ values. These results have been obtained from Latin
hypercube sampling using a sample size of 1000. (a) β1. (b) β2. (c) α1. (d) α2. (e) c1. (f ) c2.

Table 5: Summary of probabilities of existence and stability conditions of equilibria.

Equilibrium Probability of existence Probability of stability Probability of existence and stability
E1 0.5960 0.5300 0.4040
E2 0.6250 0.5370 0.4300
E∗ 0.0240 0.2190 0.001

As shown in Table 6, the transmission rate, β1, and Table 6: PRCCs for RD0 and six input parameters and the cor-
parameters c1 and c2 are highly correlated with the threshold responding sensitivity index.
RD0 with corresponding values − 0.1374 and 0.1613 and Parameters Sampling PRCCs p value Sensitivity index
− 0.2101, respectively. Moderate correlation exists between β1 LHS − 0.1374 0.1728 0.5596
transmission rates α1 and α2 and RD0 with corresponding β2 LHS 0.0025 0.9800 0.0192
values as − 0.0633 and 0.0669, respectively. Weak correla- α1 LHS − 0.0633 0.5318 0.1065
tions have been observed between the transmission rate β2 α2 LHS 0.0669 0.5085 0.5269
and RD0 with the corresponding value 0.0025. c1 LHS 0.1613 0.1090 0.7492
As shown in the sensitivity index column of Table 6, the c2 LHS − 0.2101 0.0359 0.8499
parameter c2 accounts for the maximum variability 0.8499 in
the outcome of basic reproduction number RD0. The pa-
rameter c1 is the next to account for the variability 0.7492 in
the outcome of RD0. Then, the transmission rate β1 accounts Table 7: PRCCs for RA0 and six input parameters and the cor-
for the variability 0.5596 in the outcome of RD0 followed by responding sensitivity index.
the transmission rate α2 that accounts for the variability
0.5269 in the outcome of RD0. Transmission parameters α1 Parameters Sampling PRCCs p value Sensitivity index
and β2 account for the least variability 0.1065 and 0.0192 in β1 LHS 0.2137 0.0328 0.5753
the outcome of basic reproduction number RD0, respectively. β2 LHS − 0.1928 0.0546 0.2421
Hence, parameters c1 and c2 and the transmission rate β1 are α1 LHS 0.0857 0.3967 0.0159
α2 LHS − 0.2076 0.0382 0.3872
the most influential parameters in determining RD0.
c1 LHS − 0.3343 6.7425e− 04 0.8004
It can be seen from Table 7 that the parameter c1 is highly c2 LHS 0.2523 0.0113 0.7122
correlated with the threshold RA0 with the corresponding value
12 International Journal of Differential Equations

40 40 40

30 30 30

20 20 20
RD0

RD0

RD0
10 10 10

0 0 0

–10 –10 –10

0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
(a) (b) (c)
40 40 40

30 30 30

20 20 20
RD0

RD0

RD0
10 10 10

0 0 0

–10 –10 –10

0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
(d) (e) (f )

Figure 8: Scatter plots for the basic reproduction number RD0 and six sampled input parameters’ values. These results have been obtained
from Latin hypercube sampling using a sample size of 100. (a) β1. (b) β2. (c) α1. (d) α2. (e) c1. (f ) c2.

15 15 15

10 10 10
RA0

RA0

RA0

5 5 5

0 0 0

–5 –5 –5
0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
(a) (b) (c)
Figure 9: Continued.
International Journal of Differential Equations 13

15 15 15

10 10 10
RA0

RA0

RA0
5 5 5

0 0 0

–5 –5 –5
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
(d) (e) (f )

Figure 9: Scatter plots for the basic reproduction number RA0 and six sampled input parameters’ values. These results have been obtained
from Latin hypercube sampling using a sample size of 100. (a) β1. (b) β2. (c) α1. (d) α2. (e) c1. (f ) c2.

− 0.3343. Moderate correlation exists between transmission rates number RA0 and RD0. If the threshold, RD0, is less than one,
β1, β2, and α2 and the parameter c2 and RA0 with corresponding the disagree opinion dies out and the disagree-free equi-
values as 0.2137, − 0.1928, − 0.2076, and 0.2523, respectively. librium is stable. If RD0 is greater than one, the disagree
Weak correlations have been observed between the transmis- opinion persists and the disagree-free equilibrium is un-
sion rate α1 and RA0 with the corresponding value 0.0857. stable. If the threshold, RA0, is less than one, the agree
One can see in the sensitivity index column of Table 7 opinion dies out and the agree-free equilibrium is stable. If
that the parameter c1 accounts for the maximum variability RA0 is greater than one, the agree opinion persists and the
0.8004 in the outcome of basic reproduction number RA0. disagree-free equilibrium is unstable. We simulated some
The parameter c2 is the next to account for the variability examples with different parameter values to show the ex-
0.7122 in the outcome of RA0. The transmission rate β1 istence and the stability of such equilibria.
accounts for the variability 0.5753 in the outcome of this The probabilities of the existence and probabilities of the
threshold followed by the transmission rate α2 that accounts stability of equilibria are computed based on the parameters’
for the variability 0.3872 in the outcome of RA0. Trans- distribution function sampled with the Latin hypercube
mission parameters β2 and α1 account for the least variability sampling method. To identify the most influential parameter in
0.2421 and 0.0159 in the outcome of basic reproduction the proposed model, global sensitivity analysis is carried out
number RA0, respectively. Hence, parameters c1 and c2 and based on the partial rank correlation coefficient method and
the transmission rate β1 are also the most influential pa- Latin hypercube sampling. This statistical study shows that the
rameters in determining RA0. most influential parameters in the determination of thresholds
Scatter plots comparing the basic reproduction numbers of equilibria stability are β1, the polarization parameter of
RD0 and RA0 against each of the six parameters: β1, β2, α1, α2, ignorant people by people agreeing, the parameter, c1, of the
c1, and c2 are shown in Figures 8 and 9, respectively, based loss of interest of the people agreeing, and finally c2, the pa-
on Latin hypercube sampling with a sample size of 100. rameter of loss of interest of disagreeing people.
These scatter plots clearly show the linear relationships
between outcome of RD0 and RA0 and input parameters. Data Availability
No data were used to support this study.
7. Conclusion
In this paper, an IAD model-type compartmental model has Conflicts of Interest
been considered to explore agree-disagree opinions during
polls. The equations governing the system have been solved The authors declare that they have no conflicts of interest.
to compute equilibrium states, and the next-generation
matrix method is used to derive basic reproduction numbers Acknowledgments
RA0 and RD0.
The authors would like to thank all the members of the
The model exhibits four feasible points of equilibrium,
Editorial Board who were responsible for this paper.
namely, the trivial equilibrium, agree-free equilibrium,
disagree-free equilibrium, and the positive equilibrium.
Sufficient equilibrium conditions of existence are given, and
References
stability analysis is performed to show under which con- [1] S. J. Rosenstone and J. M. Hansen, Mobilization, Participation,
ditions equilibrium states are stable or unstable. The stability and Democracy in America, Macmillan, New York, NY, USA,
of these points of equilibrium is controlled by the threshold 2003.
14 International Journal of Differential Equations

[2] L. Ezrow, C. De Vries, M. Steenbergen, and E. Edwards, Computational and Mathematical Methods in Medicine,
“Mean voter representation and partisan constituency rep- vol. 2017, 2017.
resentation: do parties respond to the mean voter position or [21] M. A. Khan, S. W. Shah, S. Ullah, and J. F. Gómez-Aguilar, “A
to their supporters?” Party Politics, vol. 17, no. 3, pp. 275–301, dynamical model of asymptomatic carrier zika virus with
2011. optimal control strategies,” Nonlinear Analysis: Real World
[3] J. Duffy and M. Tavits, “Beliefs and voting decisions: a test of Applications, vol. 50, pp. 144–170, 2019.
the pivotal voter model,” American Journal of Political Science, [22] O. Diekmann, J. A. P. Heesterbeek, and M. G. Roberts, “The
vol. 52, no. 3, pp. 603–618, 2008. construction of next-generation matrices for compartmental
[4] M. F. Meffert and T. Gschwend, “Polls, coalition signals and epidemic models,” Journal of The Royal Society Interface,
strategic voting: an experimental investigation of perceptions vol. 7, no. 47, pp. 873–885, 2010.
and effects,” European Journal of Political Research, vol. 50, [23] R. Jan, M. A. Khan, and J. F. Gómez-Aguilar, “Asymptomatic
no. 5, pp. 636–667, 2011. carriers in transmission dynamics of dengue with control
[5] R. B. Morton, D. Muller, L. Page, and B. Torgler, “Exit polls, interventions,” Optimal Control Applications and Methods,
turnout, and bandwagon voting: evidence from a natural vol. 2019, 2019.
experiment,” European Economic Review, vol. 77, pp. 65–81, [24] S. Ullah, M. A. Khan, and J. F. Gómez-Aguilar, “Mathematical
2015. formulation of hepatitis B virus with optimal control analy-
[6] T. E. Patterson, “Of polls, mountains: U.S. Journalists and sis,” Optimal Control Applications and Methods, vol. 40, no. 3,
their use of election surveys,” Public Opinion Quarterly, pp. 529–544, 2019.
vol. 69, no. 5, pp. 716–724, 2005. [25] J. C. Helton, F. J. Davis, and J. D. Johnson, “A comparison of
[7] R. Chung, The Freedom to Publish Opinion Polls: A Worldwide uncertainty and sensitivity analysis results obtained with
Update of 2012, ESOMAR, Amsterdam, Netherlands, 2012. random and Latin Hypercube sampling,” Reliability Engi-
[8] R. Lachat, “The impact of party polarization on ideological neering & System Safety, vol. 89, no. 3, pp. 305–330, 2005.
voting,” Electoral Studies, vol. 27, no. 4, pp. 687–698, 2008. [26] S. Marino, I. B. Hogue, C. J. Ray, and D. E. Kirschner, “A
[9] M. M. Pereira, “Do parties respond strategically to opinion methodology for performing global uncertainty and sensi-
polls? Evidence from campaign statements,” Electoral Studies, tivity analysis in systems biology,” Journal of Theoretical Bi-
vol. 59, pp. 78–86, 2019. ology, vol. 254, no. 1, pp. 178–196, 2008.
[10] L. A. Adamic and N. Glance, “Left Te political blogosphere [27] D. M. Hamby, A Comparison of Sensitivity Analysis Tech-
and the 2004 U.S. Election: divided they blog,,” in Proceedings niques, Westinghouse Savannah River Company, Aiken, SC,
of the 3rd International Workshop on Link Discovery USA, 1994.
(LinkKDD ’05), pp. 36–43, ACM, New York, NY, USA, [28] J. C. Thiele, W. Kurth, and V. Grimm, “Facilitating parameter
August 2005. estimation and sensitivity analysis of agent-based models: a
[11] M. D. Conover, J. Ratkiewicz, M. Francisco, B. Gonc, cookbook using NetLogo and R,” Journal of Artificial Societies
A. Flammini, and F. Menczer, “Left Political polarization on and Social Simulation, vol. 17, pp. 30–47, 2014.
twitter,” in Proceedings of the Fifh International AAAI Con-
ference on Weblogs and Social Media, pp. 89–96, Barcelona,
Catalonia, Spain, July 2011.
[12] G. Serra, “Left Should we expect primary elections to create
polarization? a robust median voter theorem with rational
parties,” in Routledge Handbook of Primary Elections,
R. G. Boatright, Ed., pp. 72–94, Routledge, New York, NY,
USA, 2018.
[13] G. Olivares, “Opinion polarization during a dichotomous
electoral process,” Complexity, vol. 2019, 2019.
[14] D. J. Pannell, “Sensitivity analysis of normative economic
models: theoretical framework and practical strategies,” Ag-
ricultural Economics, vol. 16, no. 2, pp. 139–152, 1997.
[15] https://polling.reuters.com/#!response/CP3_2/type/week/dates/
20170201-20190409/collapsed/true.
[16] M. Ludovic, Applications en Biologie. Ecologie, Environne-
ment, Université Nice Sophia Antipolis, Nice, France, 2004.
[17] O. Diekmann and A. J. Johan, “On the definition and the
computation of the basic reproduction ratio R 0 in models for
infectious diseases in heterogeneous populations,” Journal of
Mathematical Biology, vol. 28, no. 4, pp. 365–382, 1990.
[18] P. Lou, “Modelling seasonal brucellosis epidemics in bayin-
golin mongol autonomous prefecture of Xinjiang, China,
2010–2014,” BioMed Research International, vol. 2016, 2016.
[19] S. Liu, “Global analysis of a model of viral infection with latent
stage and two types of target cells,” Journal of Applied
Mathematics, vol. 2013, 2013.
[20] P. Padmanabhan and S. Padmanabhan, “Computational and
mathematical methods to estimate the basic reproduction
number and final size for single-stage and multistage pro-
gression disease models for zika with preventative measures,”