LECTURE 2 "Algorithmic models in economics and enterprise"

Summary
General provisions and basic concepts of algorithmic (simulation) modeling.
Statistical modeling method. The main stages of the simulation process. Statistical
estimates of distribution parameters. Determining the required sample size.
Plan
2.1 The essence of algorithmic (simulation) modeling
2.2 The main stages of the simulation process
2.3 Theoretical foundations of the method of statistical modeling
2.4 Modeling of random behaviors

2.1 The essence of algorithmic (simulation) modeling

As noted in the previous material, one of the classification features of
mathematical models can be classified as analytical, simulation (algorithmic) and
combined.
With the development of computer technology and discrete analysis,
algorithmic (simulation) models are becoming more widely developed and used.
Algorithmic modeling describes the process of functioning of the system in time,
simulating the elementary phenomena that make up the process, while preserving
their logical structure and sequence of flow in time.
Algorithmic models, as well as analytical ones, can be deterministic and
stochastic. In the latter case, the model uses random number sensors to simulate the
action of uncertain and random factors. This method is called the method of
statistical modeling.
To date, this method is considered the most effective method of studying
complex systems, and often the only practically available method of obtaining
information about the behavior of a hypothetical system at the design stage.

Algorithmic (simulation) modeling is a numerical method of studying systems and

processes using a modeling algorithm.
 The influence of random factors on the course of the simulated process is
simulated with the help of a specially organized draw, which is carried out with the
help of special programs called sensors or random number generators.
2.2 The main stages of the simulation process
Among the main stages of the simulation process are the following:
1) construction of a simulation model: the transition from a real object to
logical circuits that mimic its behavior, and algorithms (models);
The construction of the algorithm contains the following components:
 construction of a logical scheme of the algorithm;
 formation of mathematical relations (analytical models);
 algorithm validation.

First, as a rule, a generalized scheme of the modeling algorithm is created, which sets
the general order (course) of actions in the simulation modeling of the studied
process. After that, a detailed scheme is developed, each element of which is
transformed into an operator (group of operators) of the program.

Checking the validity of the algorithm should answer the question of how adequately
and accurately it reflects the essence of the modeled process (in a particular situation)
and the constructed conceptual model.
2) modeling of random phenomena (random events, random variables, random
functions) that are inherent in the system under study;

3) modeling of the process of system operation (based on the use of data obtained at
the previous stage) - reproduction of the process in accordance with the developed
structural scheme and formalized description (simulation runs);
4) software implementation of the simulation model: creation or adequate use
of existing software products that provide the possibility of direct practical
implementation of the model on personal computers;

5) assessment of the adequacy of the model: comparison of the results accumulated

during the experimental operation of the model, based on information obtained about
the real object being simulated, detection and analysis of discrepancies and, if
necessary, adjustments to the model;

6) conducting simulation experiments. Obviously, this stage is targeted and includes

strategic and tactical planning of experiments, own experimentation ("simulation
experiments"), which ends with the interpretation of the results and making decisions
based on the conclusions of the evaluation and management of the object (enterprise,
bank, financial firm, trade organization, holding company, etc.).

At this stage, serial calculations are performed using the program. The stage consists
of the following steps:
 planning a machine experiment;
 carrying out working calculations;
 appropriate presentation of modeling results (in tabular and graphical forms);
 providing recommendations for optimizing the functioning of the real system.

One of the important aspects in the process of work (research) with the simulation
model is the analysis of its sensitivity. It is understood as determining the degree of
variability of the values of the target indicators of the model due to the variability
(uncertainty, variability) of the initial parameters. Thus, if there is a significant
change in the simulation results with relatively small changes in the initial data, this
is a sufficient basis for additional, more detailed research, in particular, on the
relationships between the respective variables.

2.3 Theoretical foundations of the method of statistical modeling

The method of statistical modeling (or the Monte Carlo method) is a method of
studying indeterminate (stochastic) economic objects and processes, when the
internal interactions in these systems are not fully (to some extent) known.
Reproduction of a possible (accidental) state of functioning of the modeled system is
called realization (or imitation run; further - run).
After each run register a set of parameters that characterize the random event
(its implementation). The method is based on multiple runs (random
implementations) on the basis of the constructed model with the subsequent statistical
processing of the received data for the purpose of definition of numerical
characteristics of investigated object (process) in the form of statistical estimations of
its parameters.
The process of modeling the economic system is reduced to a machine
simulation of the studied process, which is modeled on a computer with all the
significant uncertainties, coincidences and the risk generated by them. Simulation is
often called simulation.
The theoretical basis of the method of statistical modeling is the law of large
numbers. In probability theory, the law of large numbers is based on proving a
number of theorems for different conditions of convergence in the probability of the
average values of the results (based on a large number of observations) to some
values.
Under the law of large numbers understand several theorems. For example, one
of the theorems of P.L. Chebyshev is formulated as follows: "With an unlimited
increase in the number of independent tests (n) the arithmetic mean free from
systematic errors and equivalent results of observations ξi- random variable ξ, which
has a finite variance D(ξ), coincides in probability to the mathematical expectation
mξ=М(ξ) of this random variable. "

This can be written as:

 n 
  i 
lim P  i 1  m    1
n 
 n 
  (2.1)

where ε – arbitrarily small positive number.

 Bernoulli's theorem is formulated as follows: "For an unlimited increase in the
number of independent attempts (n) under the same conditions, the relative frequency
m
n of occurrence of a random event 'coincides in probability to p, ie:

 m 
lim P  i  p    1
 n  (2.2)

where ε – arbitrarily small positive number.

According to this theorem to obtain the probability of a certain event, such as the
probability of states of a system pi, i=1,…,k, calculate the relative frequencies
for the number of implementations equal to n.

The results are averaged and, with some approximation, the required probabilities of
system states are obtained. The larger n, the more accurate the result of calculating
these probabilities.
Note that any statements regarding the characteristics of the simulated system
should be based on the results of appropriate tests using the methods of mathematical
statistics.
Since random events and random functions can be represented using random
variables, the simulation of random events and random functions is performed using
random variables.

The random variable is given using the distribution function, which is the probability
that the random variable X as a result of the test will take a value less than x, ie
F ( x) P ( X  x) . (2.3)

If the function F (x) is continuous and differentiable (at least piecewise), then the

continuous random variable can be given by the probability density density f (x) ,

which is derived from the distribution function F (x) .

 So,
x
dF ( x)
f ( x)  F ( x)  f ( x)dx
dx або  (2.4)
2.4 Simulation of random events
Simulation of random events is to reproduce the fact of occurrence or non-
occurrence of a random event according to a given probability.

2.4.1. Simple event simulation

Suppose there is an event A, the probability of which is equal to P (A). Using a
sensor of random numbers with a uniform distribution law on the interval (0; 1), we
choose a number  and determine the probability that  <P (A):

P ( A) P ( A)

P (  P ( A))   f ( x)dx  dx P( A).

0 0

The probability of a random variable  falling into the interval (0; P (A)) is
equal to the value of P (A). Therefore, if during the draw the number fell into this

interval, it should be assumed that event A occurred. The opposite event ( A ) will
occur with probability (1-P (A)) in the case when   P (A).
The procedure for modeling a simple event in a simulation model is described
by an algorithm, the scheme of which is given in Fig. 3.1 (GVC () - generator of
random numbers  corresponding to the uniform distribution law on the interval (0;
1)).
 Figure 2.1 - Simulation of a simple event
Operator 1 refers to a random number generator, which generates a random
number .
Operator 2 checks the condition   P (A). If it is executed, it is considered that
event A has occurred. Otherwise, it is considered that the opposite event has occurred
( A ).

EXAMPLE
The probability of occurrence of event A in each test is equal to P (A) = 0.75.
It is necessary to simulate three tests and determine the sequence of implementation
of event A.
SOLUTION.
We set aside a point E = 0.75 on a unit segment of the numerical axis and
assume that when the random number ξi <E, then in the test came the event A.
Otherwise (at ξi ≥ Е) event occurred not A ( A ), ie event A did not take place.
Assume, for example, that random numbers ξ1 = 0.925, ξ2 = 0.135, ξ3 = 0.088 are
selected from the corresponding table. Then for three tests we get the following
sequence of events: A , A, A.

2.4.2. Simulation of a complete group of pairwise incompatible events Suppose that

there is a complete group of pairwise incompatible random events A1, A2,…, Ak
with probabilities p1, p2,…, pk. The condition is fulfilled:
k

p
i 1
i 1.

Divide the interval (0; 1) into k segments whose lengths are respectively equal
to p1, p2,…, pk.

p1 p2 p k–1 pk

0 L1 L2 Lk–1 Lk =1 ξ

Figure 2.2 - Modeling of the full group of porn incompatible divisions

 Lj
j
L j  p i , P ( L j  1   L j )  dx L j  L j  1 p j .
If we denote i 1 , it is a fair equality Lj 1

Therefore, if the random variable  generated by the sensor of random numbers

that correspond to the uniform distribution law on the interval (0; 1) falls, for
L j  1; L j
example, on the interval , it means that the event Aj has occurred.

The procedure for modeling a complete group of pairwise incompatible events is

described by an algorithm, the scheme of which is shown in Fig. 3.3.

Figure 2.3 - Scheme of the algorithm for modeling a complete group of

pairwise incompatible events

Operator 1 refers to the generator of random numbers corresponding to the

uniform distribution law on the interval (0; 1). Operator 2 checks the condition of the
random variable  in the interval (0; L1). If this condition is met, it is considered that
event A1 has occurred. If this condition is not met, the algorithm involves checking
the conditions for the random variable to fall into other intervals.

2.4.3. Simulation of compatible (dependent and independent) events. Simulation of

compatible events can be performed in two ways. 1 way. Identify all possible variants
of occurrence of compatible events in the test, find a complete group of pairwise
incompatible events and calculate their probabilities. Then do the same as in the
simulation of a complete group of incompatible events.

Example. Suppose that dependent and compatible events A and B take place
during the test, and it is known that P (A) = 0.7; P (B) = 0.5; P (AB) = 0.3.

It is necessary to simulate the occurrence of events A and B in two trials.

Solution. In each test, the occurrence of one of four pairwise incompatible
events is possible:
С1 = АВ, Р(С1) = Р(АВ) = 0,3.
С2 = AB , Р(С2) = Р( AB ) = Р(А) – Р(ВА) = 0,7 – 0,3 = 0,4.

С3 = A B , Р(С3) = Р( A B ) = Р(В) – Р(АВ) = 0,5 – 0,3 = 0,2.

С4 = A B , Р(С4) = 1 – [Р(С1) + Р(С2) + Р(С3)] = 1 – (0,3 +
+ 0,4 + 0,2) = 0,1.
We model the complete group of events C1, C2, C3, C4 in two tests (runs). Set
aside sequentially on a single segment of the numerical axis intervals (Fig. 1):
і = Р(Сі), і = 1,…, 4.

Δ1 = Р(С1) Δ2 = Р(С2) Δ3 = Р(С3) Δ4 = Р(С4)

0 0,3 0,7 0,9 1

Figure 2.4 – Intervals і = Р(Сі)

Let two random numbers 1 = 0.68 and 2 = 0.95 be generated. The random number
1 belongs to the interval 2, so in the first test, event C2 took place: event A
occurred, and event B did not occur. In the second test, the random number 2
belongs to the interval 4: both events A and B did not occur.
2 way. Simulation of compatible events is to act out the fact of occurrence of
each of the compatible events separately, and if the events are dependent, it is
necessary to pre-determine the conditional probabilities.
 Example. Using the conditions of the previous example, it is necessary to
simulate the occurrence of events A and B in one test.
Solution. Events A and B are dependent, so we first find the conditional
probabilities Р(В/А) та Р( B / A ):
P( AB) 0,3 3
P ( B / A)    ;
P( A) 0,7 7
P( A B) 0,2 2
P( B / A )    .
P ( A ) 1  0,7 3
A random number 1 was chosen to simulate event A. Let 1 = 0.96. Since 1>
P (A), event A did not occur in the test.
Now let's play event B, provided that event A did not take place in the test. Let

the random number 2 = 0.22. Therefore  2  P( B / A ) , (0.22 <2/3), ie event B in the

test occurred.

2.4.4 Modeling of a discrete random variable

The distribution of the discrete random variable X can be represented as a
table:
xi x1 x2 … xn
pi p1 p2 … pn
Here pj- is the probability that the random variable X acquires the value xj, ie
p j  P X  x j 
, j = 1,…, n. The condition is fulfilled
n

p j 1
j 1
.
Divide the interval (0; 1) into n segments whose lengths are equal to the given
probabilities pj, j = 1,…, n. If a random number  formed by a generator of random
numbers that correspond to a uniform distribution law in the interval (0; 1) falls into
the interval pk, then the random variable X becomes xk. Thus, when modeling a
discrete random variable, the same procedure is actually used as when modeling a
complete group of pairwise incompatible events.
Example. Model two implementations of a discrete random variable X with
distribution:
 x -1 0 2 5
i:
p 0, 0 0,2 0,1
i: 2 ,5
Solution. Set aside sequentially on a single segment of the numerical axis
intervals (Fig. 1): і = pi, і = 1,…, 4.

Δ1 = p 1 Δ2 = p 2 Δ3 = p 3 Δ4 = p 4

0 0,2 0,7 0,9 1

Figure 2.5 – Intervals і = pі

Let two random numbers be generated 1 = 0,57 і 2 = 0,73.
The random number 1 belongs to the interval 2, so in the first test the
random variable X becomes x2 = 0. In the second test the random number 2 belongs
to the interval 3, so the random variable X takes the value x3 = 2.
2.4.5 Simulation of a continuous random variable
Modeling of random variables with uniform distribution. The random number
generator generates a sequence of realizations of the random variable ξ, with a
uniform distribution function on the interval (0; 1). For the most part, it is necessary
to model random variables with a uniform distribution, which acquire values in the
interval (a; b).
Suppose that
x a
  F ( x) 
b a
from here
x a   (b  a )

In practice, a slightly modified method is used. Instead of the boundaries of the

interval are given: the average value of the random variable m (ξ) and the value
(length) of the interval Δx, within which this random variable can acquire its values
(with a uniform distribution law). Determining the possible value (realization) of a
random variable with a uniform distribution can be done according to the expression:
x m( )  x(  0,5)
 Modeling of random variables with normal distribution law. According to the
central limit theorem of probability theory, due to the addition of a sufficiently large
number of equally distributed independent random variables, we obtain a random
variable that has a normal distribution law.
Studies have shown that as a result of adding more than ten random
independent quantities with a uniform distribution in the interval (0; 1) we obtain a
random variable, which with accuracy sufficient for most practical problems can be
considered distributed according to normal law.
The procedure for drawing a normally distributed random variable:
1) add, for example, 12 independent random variables that have a uniform
distribution law and which acquire values in the interval (0; 1), ie:
12
   i
i 1

Using the known theorems on the mathematical expectation of the sum of

independent random variables with the same distribution law and variance of this
sum, we can establish that the random variable ν has the following characteristics:
12
1
m( )  m( i ) 12  6
mathematical expectation: i 1  2
12
 1
D ( )  D( i ) 12  1
variance D (v): i 1  12 

standard deviation  ( )  D( ) 1 .

B. Normalize and center the random variable ν, ie move to a random variable 

that has zero mathematical expectation and standard deviation    1 .


  m    6
  
From the normalized and centered random variable, you can go to the random
variable y with the given parameters т(у) and  (у) according to the following
expression:
y m( y )  ( y )

Simulation of a random variable with a given integral distribution function. To

convert a sequence of random numbers that are realizations of a random variable with
a uniform distribution law in the interval (0; 1) into a sequence of random numbers
that are realizations of a random variable with a given integral distribution function F
(x), it is necessary from a set of random numbers with a uniform the law of
distribution in the interval (0; 1) choose a random number ξ and solve the equation:
F(x) = ξ relatively х. (2.5)
In the case when the probability density function f(x) is given, the relation (2.5)
takes the form:

f x dx 
 (2.6)
For a number of distribution laws, an analytical solution of equation (3.6) is
obtained, the result of which is given in Table 3.1.
Table 2.1 - Formulas for modeling random variables
Laws of distribution of Formulas for modeling
Density of distribution
a random variable random

Exponential f x  e  x xi 
1
ln  i

Weibula xi  bln  i 
a 1 1/ a
a x   x a 
f x     exp     
ba   b  


Gamma distribution   x   1 1
f x  
 
e x xi 

 ln(1  
j 1
ij )
(η - integers)

Normal 1
 x  m 2  12 
f x   e 2 2 xi m      ij  6 
 2  j 1 

Modeling of random variables with interval-constant distribution function.

Suppose there are grounds to approximate the distribution function of a random
variable X, which is given on the interval [a0; an], an interval-constant distribution
density function f(x). This means that the segment [a0; an] is divided into n segments
so that the probabilities of hitting each of them are known (pk, k = 0, 1, …, n) (Fig.
2.4).
 f(x)

a0 a1 a2 an–1 an х

Figure 2.4 - Interval-constant density function

distribution of a random variable
That is
ak

f ( x)dx p
ak  1
k
, k 1, ..., n .

From the condition f(x)=const=ck on each partial interval it follows that the
implementation of a random variable X can be determined by the formula

x k a k  1   (a k  a k  1 ), k 1, ..., n ,
(2.7)
where  is the realization of a random variable evenly distributed on the
interval (0; 1); ak–1 - left boundary of the partial interval; ak is the right limit of the
partial interval.

Hitting any partial interval can be considered as an event that is part of a

complete group of pairwise incompatible events, and the number of the
corresponding interval - as a discrete random variable  with the distribution:
i 1 2 … n
pi p1 p2 … pn

Therefore, the modeling procedure in general is as follows:

1. Using a random number generator we model a discrete random variable  -
the number of the interval.
2. Using a random number generator, play the random variable ξ (with a
uniform distribution on the interval (0; 1)) and determine the implementation of the
random variable X by formula (9.3). The block diagram of the algorithm is shown in
Fig. 3.5.

Figure 3.5 - Block diagram of a random simulation algorithm values with

interval-constant distribution density function

Example. Model two implementations of a random variable X having an

interval-constant distribution density function on a segment [0,5; 7,5]:
a0 0,5; a1 2,0; a2 4,5; a3 6,5; a4 7,5

p1 0,3; p2 0,4; p3 0,2; p4 0,1

Solution. For a given random variable X, the discrete random variable 

corresponding to the interval number will have a distribution:
i: 1 2 3 4
pi: 0,3 0,4 0,2 0,1
For one implementation of a random variable X it is necessary to generate one
value (implementation) of a discrete random variable  (see previous example) and
one random number  formed by a generator of random numbers that correspond to a
uniform distribution law on the interval (0; 1).
Suppose that two realizations of a discrete random variable are generated: :
1 = 2 2 = 4 , and two random numbers 1 = 0,91, 2 = 0,43.
 Then the first value (implementation) of the random variable X will belong to
the second interval: 2,0 ; 4,5 , and substituting the values of the corresponding
x1 a1  1 (a2  a1 ) 2,0  0,91(4,5  2,0) 4,275
quantities in the formula (2.1.3) , we obtain: .
Similarly, the second value (implementation) of the random variable X will
belong to the fourth interval: 6,5 ; 7,5 , and substituting the values of the corresponding
quantities in the formula (2.1.3), we obtain:
x2 a3   2 (a4  a3 ) 6,5  0,43(7,5  6,5) 6,93 .

An example of building a simulation model

The entrepreneur is going to invest in the construction of a new enterprise that
will produce certain products that are in demand in the market. Similar products are
produced by other companies, so we will have to operate in a competitive
environment.
It is possible to estimate, ie to consider known, the mathematical expectation
(or mean value) and the standard deviation of the value of operating costs of
production. You can also accept the hypothesis that the costs will have a normal
distribution law with given parameters.
It is assumed that the market capacity as a random variable has, for example,
also a normal distribution law with known parameters (mathematical expectation and
standard deviation).
It is more difficult to determine the characteristics of the market share (as a
random variable), which can occupy this company after its commissioning. Assume
that you can predict only the average market share. The type of distribution is mostly
unknown, and there are insufficient grounds to consider the distribution to be normal.
In this case, it is appropriate to use distributions from another class (for example,
uniform or interval-uniform). It is advisable to consider several distribution options
and analyze the model's response to changes in both selected distribution functions
and their parameters.
For the indicator of efficiency of the enterprise it is appropriate to choose, in
particular, the profit from the sale of products and to estimate the value of the
guaranteed profit for a given degree of one of the quantitative indicators of risk. If
there are grounds for this, then the hypothesis is accepted that the random value of
profit has a normal distribution law.
So, let's form a conceptual model:
Production is associated with operating costs, which (hypothetically) are a
random variable (Rrach) with a normal distribution law with given parameters: the
mathematical expectation mrach and the standard deviation rach.
The market capacity where the company's products are to be sold is also a
random variable (Rryn), which has (assumption) a normal distribution law with given
parameters: the mathematical expectation of mryn and the standard deviation of ryn.
The market share of the enterprise is uncertain and can be given by some
random variable (dryn) with a certain distribution function (for example, interval-
uniform function).
We assume that the company's profit is a random variable (Rprof), which is
determined from the expression:
R prof Rryn d ryn  Rrach ,
where Rprof is the random value of the company's profit; Rryn - a random
variable of market capacity; dryn - random value of the market share of the
enterprise; Rrach - a random variable of operating costs of the enterprise.
The resulting characteristics of the model will be considered:
i
R prof
• the sum of the values of the random variable Rprof for N
implementations (simulation runs):
N
S prof  R prof
i
;
i 1

R iprof
• the sum of the squares of the values of the random profit for N
implementations (simulation runs):

 R prof .
N
2 i 2
S prof
i 1

An indicator of the efficiency of the enterprise will choose a guaranteed profit

for a given level of risk, which will be determined by the formula
G prof m prof  k   prof ,
 where Gprof is the guaranteed amount of profit according to the set value of
the risk indicator ; mprof - estimate of the mathematical expectation of a random
S prof
m prof 
variable: N prof - estimate of the standard deviation of a random profit:
1
 prof  2
( S prof  N m 2prof ) ;
N1
k - some coefficient that depends on  and is
determined by the distribution function of the random variable Rprof.
If there is reason to accept the hypothesis of the normal distribution law of the
random variable Rprof, then k - is the quantile of the normal distribution law
according to the given value of the component  of the risk vector, for example, if 
= 0,1, то k = 1, (determined by tables integral function of the standard normal
distribution law).
The general view (layout) of the starting form (as a sample) is shown in Fig.
3.6.
The layout includes the following controls: multiple labels with object names;
fields for adjusting input data, as well as fields for outputting simulation results.

Figure 2.6 - Layout of the starting form

 The generalized scheme of the algorithm for calculating the value of Gprof is
shown in Fig. 3.7.
Operator 1 enters the input data and all the necessary parameters.
Operator 2 is the beginning of the cycle of simulation runs (implementations).
The cycle ends when I = N.
Operator 3 refers to a procedure that generates a possible value of a random
variable that is distributed according to the standard normal distribution law.
Operator 4 calculates the implementation of a random amount of operating
costs.
Operators 5 and 6, similarly to the previous one, determine the implementation
of a random variable of market capacity.
Operator 7 refers to the procedure that determines the realization of the random
value of the company's market share. This random variable is generated, in particular,
according to the procedure for modeling random variables with interval-uniform
distribution.
Operator 8 determines the realization of a random amount of profit.
Operator 9 accumulates the amount of sales of random profit and the sum of
their squares for all runs.
After completing the simulation runs, the operator 10 calculates the values
mprof, prof, Gprof .
Operator 11 displays the simulation results on a monitor (or printer).
 Figure 2.7 - Generalized algorithm for calculating the value Gprof
Numerical example.
Let the following data be known:
mrach - average value of operating costs (UAH): mrach = 110 000;
rach - standard deviation of operating costs: rach = 11 000;
mryn - the average value of market capacity: mryn = 2,780,000;
ryn - standard deviation of market capacity: ryn = 250 000;
N - the number of random implementations: N = 1000.
We will consider the parameters of the law of distribution of the enterprise's
share in the market as variable parameters.
Consider three distribution options.
 For the first option, assume that the distribution law is uniform, ie the number
of boundary points n = 2. Let the average value of market share is 0.1, and the value
of the coordinates of the boundary points: а0 = 0,099; а1 = 0,101. Therefore, for the
first option, the degree of uncertainty is quite small. The company's market share is
almost constant (10% of its total capacity).
For the second option, assume that the number of boundary points п = 6 (the
range consists of five intervals), ie the random value of market share is distributed
unevenly. Let the mean value be the same (0,1) and the boundary points be
symmetrical with respect to the mean value. Choose the following values of the
coordinates of these points:
a 0 0,035; a1 0,075; a 2 0,095; a 3 0,105; a 4 0,125; a 5 0,165.
The probabilities of hitting individual intervals will be determined from the
1 1
pi   , i 1,  ,5
condition that they are all the same and equal to: n 1 5 .
Then it is easy to find the corresponding values of the distribution density at
each of the five intervals (Fig. 3.8):
f 1 5; f 2 10; f 3 20; f 4 10; f 5 5.

fryn
f3 = 20

10 f2 = 10 f4 = 10

f1 = 5 f5 = 5
5

0,1 0,2 d ryn

Figure 2.8 - Interval-uniform law of market share distribution (second option)

For the third variant, it is known that the number of boundary points is п = 6
(the range consists of five intervals), but the boundary points are asymmetric with
respect to the mathematical expectation. Let the average value of this random
variable be approximately 0.1, and the coordinates of the boundary points:
a 0 0,035; a1 0,075; a 2 0,095; a 3 0,105; a 4 0,155; a 5 0,255.
The values of the distribution density at each of the five intervals are,
respectively:
 f 1 5; f 2 10; f 3 20; f 4 4; f 5 2.
Data on these three options for the company's market share can be presented in
the table (Table 2.2).

Table 2.2 - Parameters of interval-uniform distributions

Number Coordinates of points
Number option
points
1 2 3 4 5 6
1 2 0,099 0,101 — — — —
2 6 0,035 0,075 0,095 0,105 0,125 0,165
3 6 0,035 0,075 0,095 0,105 0,155 0,255

The simulation results are given in table. 2.3

Table 2.3 - Simulation results
Option number mprof σprof Gprof

1 168822,62 27440,84 133698,34

2 174401,02 91653,35 57084,72
3 216211,40 152847,04 20567,19
Data analysis table. 2.3 shows that with increasing uncertainty and the resulting
degree of risk in relation to the company's market share, the guaranteed profit
decreases due to a larger spread (variation) of the random amount of profit.
If there is no additional information, the first option is better. Other risk
decision strategies are possible.