Vol.

9, 2024-03

Constrained Randomistic Variables

Hugo Hernandez
ForsChem Research, 050030 Medellin, Colombia
[email protected]

doi: 10.13140/RG.2.2.12411.73764

Abstract
Randomistic variables integrate the realms of deterministic and random variables. Randomistic
variables are represented by probability distribution functions, and in the case of continuous
variables, also by probability density functions (just like random variables). Any randomistic
variable can be subject to external constraints on its possible values. Thus, the resulting
probability distribution of the constrained variable may be different from the probability
distribution of the original variable. In this report, general expressions for analytically
determining the probability distribution functions (or probability density functions) of
constrained randomistic variables are presented. These expressions are extended to
constraints involving multiple, independent randomistic variables. Several illustrative examples,
with different degrees of difficulty, are included. These examples show that constrained
randomistic variables represent the solution to a wide variety of problems, including algebraic
systems of equations, inequalities, magic squares, etc. Further improvements in analytical and
numerical methods for finding constrained probability functions would be highly desirable.

Keywords
Constraints, Equations, Inequalities, Membership Function, Multivariate, Probability Density,
Probability Distribution, Randomistics, Sets

1. Introduction

The term randomistic represents the integration of the random and deterministic domains [1].
In general, a quantitative randomistic variable can be described by the following expression:

(1.1)
where represents the magnitude of the deterministic term (bias), represents the
magnitude of the pure random term (uncertainty), represents the standard deterministic

Cite as: Hernandez, H. (2024). Constrained Randomistic Variables. ForsChem Research Reports, 9, 2024-
03, 1 - 31. Publication Date: 07/03/2024.
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

variable (with mean value and variance , and corresponding to number ), and represents
a type I standard random variable (with mean value and variance ) [1].

The frequency of observation of each possible value of a randomistic variable can be

represented by the probability distribution function ( ), corresponding to the probability
that the randomistic variable has an exact value of :

( ) ( )
(1.2)
The probability distribution function can also be expressed in terms of a probability density
function ( ( ) ) as follows:

( ) ( )
(1.3)
where represents the difference between and the next possible value of the randomistic
variable. For continuous (or approximately continuous [2]) variables, , and therefore:

( ) ( )
(1.4)
The probability distribution function of real randomistic variables has two important
conditions:
( )
(1.5)

∑ ( )

(1.6)
For real continuous variables, these conditions can be expressed as follows:

( )
(1.7)

∫ ( )

(1.8)
Particularly, the probability distribution function of the standard deterministic variable ( ) can
be expressed as the Dirac measure ( ):

( ) ( ) {
(1.9)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (2 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Similarly, the probability density function of the standard deterministic variable ( ) can be
described in terms of Dirac’s delta function ( ) as follows1:

( ) ( ) {

(1.10)

On the other hand, a probability distribution function for any type I standard random variable
( ) must fulfill the following conditions, in addition to Eq. (1.5) and (1.6):

∑ ( )

(1.11)

∑ ( )

(1.12)
Similarly, the probability density function of any type I standard random variable ( ) must fulfill
the following conditions, in addition to Eq. (1.7) and (1.8):

∫ ( )

(1.13)

∫ ( )

(1.14)
Now, considering the change of variable theorem [3], the probability distribution and
probability density functions of a randomistic variable can be expressed in terms of the
probability functions of the corresponding standard deterministic and random variables as
follows2:
( ) ( ) ( ) ( )
(1.15)

( )
( ) ∫ ( ) | |
( )

(1.16)

1
Notice that ( ) ( ) .
2
It is assumed that the standard deterministic ( ) and the standard random ( ) variables are completely
independent from each other.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (3 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

We may say that a randomistic variable is “known” when the probability distribution function
of the variable is known.

2. Constrained Randomistic Variables

2.1. Effect of Constraints on Probability Distribution Functions

When external constraints are imposed on the possible values of the randomistic variable, the
probability distribution function of the variable may change [4].

For example, let us consider that one or more external constraints are present, such that the
possible values of the variable are now limited to the set , characterized by the following
membership function:

( ) {
(2.1)

The probability that the randomistic variable satisfies the constraint will be given by [4]:

( ) ∑ ( ) ( )

(2.2)
and for continuous variables:

( ) ∫ ( ) ( )

(2.3)

Then, the probability distribution of the constrained randomistic variable ( ( ) ( ))

will be:
( )
( ) ( )
( )
(2.4)
For continuous variables, the probability density function of the constrained variable becomes
( ( ) ( ) ):
( )
( ) ( )
( )
(2.5)
( )
The term will be denoted as a constraint correction factor.
( )

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (4 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Clearly, the key to finding the probability distribution function of the constrained variable lies in
determining the constraint correction factor for each possible value of the randomistic
variable.

2.2. Multiple Constraints

Let us assume that different constraints ( ) are simultaneously imposed on the

randomistic variable . Each constraint can be represented by its own membership function
( ):

( ) {
(2.6)

Typically, multiple constraints represent joint constraints which must be all simultaneously
fulfilled. In this case, the overall membership function ( ) for the set of constraints can be
determined as follows:

( ) ∏ ( )

(2.7)
representing the intersection of multiple sets of values [4].

Notice that ( ) only when ( ) for . If at least one of the constraints has
a (Boolean) membership function of (zero), then the overall membership function is also
zero.
Nevertheless, certain constraints can be disjunctively imposed on the variable, implying that at
least one of the constraints must be fulfilled. In the case of disjunctive constraints, the overall
membership function can be determined as follows:

( ) ( ( ))
(2.8)
In the case of two disjunctive constraints, the overall membership function can be alternatively
expressed as [4]:

( ) ( ) ( ) ( ) ( )
(2.9)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (5 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

2.3. Types of Constraints

A constraint simply represents an arbitrary set of values. The most direct way to describe a
constraint is by listing the elements of the set, like for example (considering elements):

{ }
(2.10)
The elements contained by the constraint might be described as the solution of an
equation (equality relation or equality constraint), as follows:

( )
(2.11)
where ( ) represents an arbitrary function of , such that the inverse of the function yields the
elements of the constraint:
( )
{ }
(2.12)
The simplest type of equation is the linear constraint:
( )
(2.13)
In this case, the set comprises only a single value:
{ }
(2.14)
or alternatively,

(2.15)
As an additional example, let us consider the following equality constraint:
( )
(2.16)
The set of values described by the constraint is now:
{ }
(2.17)
alternatively expressed as:

(2.18)
Another type of constraints is described by inequalities. Two basic types of inequalities are
possible:

 Less-than inequality ( ):
( )
(2.19)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (6 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

 Greater-than inequality ( ):

( )
(2.20)
However, notice that multiplying both sides of Eq. (2.19) by -1, we obtain:

( )
(2.21)
And we may conclude that the same constraint can be expressed as a less-than equality or a
greater-than equality simply by considering:

( ) ( )
(2.22)
In some cases, multiple constraints can be integrated in a single expression. Some examples
include:

 Not equal to ( ):

( ) ( ) ( )
(2.23)

 Less than or equal to ( ):

( ) ( ) ( )
(2.24)
 Greater than or equal to ( ):

( ) ( ) ( )
(2.25)
 Open interval:

( ) ( ) ( )
(2.26)
 Closed interval:

( ) ( ) ( )
(2.27)
 Half-open intervals:

( ) ( ) ( )
(2.28)
( ) ( ) ( )
(2.29)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (7 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

2.4. Empty Constraint Sets

In certain cases, it is possible that the constraints imposed on the variable result in an empty
set, as follows:

(2.30)
The corresponding constraint membership function will be:

( )
(2.31)
with constraint set probability (from Eq. 2.2):

( )
(2.32)
( )
Since the calculated constraint correction factor is, for this case, indetermined ( ), the
( )

probability distribution functions of the constrained variable are also indetermined. However,
for practical purposes, let assume the following probability distribution functions3:

( )
(2.33)
( )
(2.34)
It is also important to remark that an empty constraint set is different from a constraint set
with a single value of zero (which is not an empty constraint set).

2.5. Single-Value Constraint Sets

Let us now assume that the continuous variable is subject to the following constraint:
{ }
(2.35)
where is a constant value.
The corresponding constraint membership function is (from Eq. 2.1):

( ) ( ) {
(2.36)
with constraint probability (Eq. 2.3):

3
Notice however that these functions do not satisfy the conditions given in Eq. (1.6) and (1.8).

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (8 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( ) ∫ ( ) ( ) ( ) ( )

(2.37)
Now, the probability density function of the constrained variable is (from Eq. 2.5):

( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) {
( ) ( ) ( )

(2.38)
representing the probability density of a deterministic variable having an exact value of . In
other words,

(2.39)
In general, the result shown in Eq. (2.38) is obtained for any arbitrary probability density
function of , as long as ( ) . The case where ( ) results in indetermination,
equivalent to the case of an empty constraint set (Section 2.4).

2.6. Multiple-Value Constraint Sets

The most general type of constraint set is the multiple-value set, expressed as follows:

{ }
(2.40)
where is the number of elements in the constraint set .
In this case, each possible value has a membership function to the constraint set :

( ) ∑ ( ) {

(2.41)
The resulting constrained probability distribution function ( ) fulfills:

( )
(2.42)
( )
(2.43)

∑ ( ) ∑ ( )

(2.44)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (9 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Assuming that the randomistic variable can be treated as a continuous variable, then the
corresponding probability density function of the constrained variable must fulfill the following
conditions:
( )
(2.45)
( )
(2.46)

∫ ( ) ∑ ( )

(2.47)
The number of elements in the constraint set can be considered finite (discrete) or infinite
(continuous). Of course, for discrete randomistic variables the number of elements in the
constraint can only be discrete.

Particularly in the case of discrete constraint sets, the probability distribution function and the
probability density function can be expressed in terms of Dirac’s functions as follows (from Eq.
2.4, 2.5 and 2.41):
∑ ( ) ( )
( )
∑ ( )
(2.48)

∑ ( ) ( ) ∑ ( ) ( )
( )
∑ ( ) ∑ ( )
(2.49)

2.7. Probability Functions as Constraints

So far, constraint sets were described by different relations (equalities or inequalities)

associated to functions of the randomistic variable . Now, we will consider a more general
representation of constraints based on probability functions [5].

For example, the equality constraint given in Eq. (2.11) can be alternatively expressed as
follows:
(( ) ) ( ( ) )
(2.50)
where ( ) indicates the conditional probability of statement given statement .

Eq. (2.50) can then be further transformed into:

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (10 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( ( ) ) ( )
(2.51)
where ( ) is now a function involving probability functions.

On the other hand, the inequality constraint given in Eq. (2.19) will now be expressed as
follows:
( ( ) )
(2.52)
or alternatively,
( ( ) )
(2.53)
Notice that in this case, an inequality constraint is transformed into an equality constraint of
probability functions.

Multiple constraints can also be represented in terms of probability functions. For example, let
us consider two conjunctive constraints expressed as in terms of the randomistic variable as
follows:
( )

( )
(2.54)
Can be expressed as a single constraint in terms of probability functions as follows:

( ( ) ) ( ( ) )
(2.55)
If the constraints considered in Eq. (2.54) were disjunctive, the resulting overall constraint
would be:

( ( ) ) ( ( ) ) ( ( ) ) ( ( ) )
(2.56)
Inequality constraints of probability functions are also possible. For example, the overall
constraint for two disjunctive constraints can alternatively be expressed as follows:

( ( ) ) ( ( ) )
(2.57)
However, when an equivalent equality expression is available, the equality will be preferred.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (11 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

2.8. Multivariate Constraints

Up to this point, constraints influencing only a single randomistic variable were considered.
However, in a more general scenario, constraints may involve two or more independent
randomistic variables ( ) simultaneously.

In terms of simple functions of the randomistic variables, we may represent equality and
inequality constraints as follows:
( )

(2.58)
( )

(2.59)
It is also possible to describe constraints in terms of probability functions.

However, the resulting multidimensional constraint set cannot be easily represented, as we do

not have a set of feasible values but a set of feasible combinations of values.

The constraint set membership is now represented as follows:

( )
{

(2.60)
and the constraint set probability will be given by:

( )
∑ ∑ ∑ ∏
( ) ( )

(2.61)
and for continuous variables:

( )
∫ ∫ ∫ ( )
∏( ( )
)

(2.62)
The main difficulty lies in the analytic determination of the constrained probability distribution
functions for individual randomistic variables. For example, for the first randomistic variable
( ) we obtain:
∑ ∑ ∏
( ) ( )

( )
( )

(2.63)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (12 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

And for continuous variables:

∫ ∫ ( ) ( )
∏ ( ( )
)
( )
( )

(2.64)

3. Illustrative Examples

In this Section, some illustrative examples are used to provide a better understanding of the
concepts introduced in Section 2. In all cases, the goal is finding the constrained probability
distribution function or the constrained probability density function of the randomistic
variable(s) subject to the constraint(s) considered.

3.1. Example # 1: Single-Value Constraint Set imposed on a Standard Normal Random Variable

As a first example, let us consider a continuous randomistic variable given by the following
expression:

(3.1)
where is a standard normal random variable with probability density:

( )
√
(3.2)
Thus, from Eq. (1.16), and considering that and :

( ) ( )
√
(3.3)
Now, let us assume that this variable is subject to the following equality constraint:

(3.4)
with membership function:

( ) {
(3.5)

The constrained probability density function then becomes (from Eq. 2.5):

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (13 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( )
( ) ( ) {
( )

(3.6)
Corresponding to Dirac’s delta function ( ). It means that the solution to this problem is
simply:

(3.7)
The original and constrained probability density functions obtained for this example are
depicted in Figure 1.

Figure 1. Original (solid blue line) and constrained (dashed red line) probability density function for
Example # 1.

3.2. Example # 2: Same Single-Value Constraint Set but now imposed on a Standard Uniform
Random Variable

The previous example may appear quite trivial since Eq. (3.7) can be directly obtained by
rearranging Eq. (3.4); however, let us now assume that the same constraint was imposed on
the standard uniform random variable:

(3.8)
with probability density function:

( ) ( ) {
(3.9)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (14 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

The constrained probability density function now is (from Eq. 2.5):

( )
( ) ( ) {
( )

(3.10)
The result in this case is indetermined, indicating that there is no solution since an empty set is
obtained.
Thus, even when a value of the variable can be obtained directly from the constraints (like Eq.
3.7), if it is an impossible result for the original variable, then it is not a valid solution to the
problem.
Figure 2 illustrates the original and constrained probability density functions obtained for this
example. The constraint solution ( ) is included for comparison.

Figure 2. Original (solid blue line) and constrained (dashed red line) probability density function for
Example # 2. The solution to the constraint is represented by the dashed gray line.

3.3. Example # 3: Non-Uniform Multiple-Value Constraint

Let us consider again the standard normal random variable introduced in Eq. (3.1), with
probability density function described by Eq. (3.3). The constraint imposed in this case is the
following:

(3.11)
As a result, the constraint set is represented by only two elements:

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (15 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

{ }
(3.12)
and with membership function:
{ }
( ) {
{ }
(3.13)

Thus, the constrained probability density function obtained is:

{ }

( )
( )
( ) ( )
( )

( )
{
(3.14)
which can be alternatively expressed as follows:

( ) ( )
( )

(3.15)

Figure 3. Original probability density function (solid blue line) and constrained probability function
(dashed red line) for Example # 3.

In terms of probabilities, we have:

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (16 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( ) ( )
( ) ( )

(3.16)
This means that two possible values can solve the problem: with a probability

( ( ) ), and with probability ( ( ) ).

Figure 3 shows the probability distributions obtained for this problem.

Notice that when the original probability distribution is uniform, all feasible values obtained in
the constraint set will have identical probabilities.

3.4. Example # 4: Inequality Constraint imposed on the Standard Normal Random Variable

Let us continue considering the standard normal random variable introduced in Eq. (3.1), with
probability density function described by Eq. (3.3). The inequality constraint imposed in this
case is the following:

(3.17)
with membership function:

( ) {
(3.18)

Figure 4. Original (solid blue line) and constrained (dashed red line) probability density functions for
Example # 4.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (17 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

The probability density function obtained for the constrained variable is:

( )
( ) ( ) √
( ) √
∫ ( )

(3.19)
and ( ) for .

Figure 4 illustrates the probability density functions for the original and constrained variables.

3.5. Example # 5: Periodic Constraint imposed on the Standard Normal Random Variable

This example again considers the standard normal random variable as the original variable. The
constraint imposed in this case is the following:

( )
(3.20)
The constraint set obtained can be represented by the following expression:

( )

(3.21)
or in terms of the periodic equality operator [6]:

(3.22)
The corresponding membership function is:

( ) {

(3.23)

Now, the constrained probability density function becomes:

( )
( ∑ )

(3.24)
with probabilities:

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (18 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( )
∑
(3.25)
Figure 5 shows the original probability density function along the constrained probability
distribution, limited to the range .

Figure 5. Original probability density function (solid blue line) and constrained probability function
(dashed red line) for Example # 5.

3.6. Example # 6: Multivariate Equality Constraint imposed on two Uniform Variables

Let us consider now a multivariate problem, involving two independent randomistic variables
having uniform distributions, but different ranges as follows:

(3.26)

Variable is uniformly distributed in the range , whereas variable is uniformly

distributed in the range . Notice that , indicating that the variables are
independent from each other.

The probability density functions of the original variables are:

( )
{

(3.27)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (19 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( )
{

(3.28)
Let us now consider the following constraint imposed on these variables:

(3.29)
with membership function:

( ) ( )
{

(3.30)
4
The probability of the constraint set is (from Eq. 2.62) :

( )
∫ ∫ ∫ ∫
( )

(3.31)
The first integral is limited to because (from Eq. 3.29 and 3.28):

(3.32)
and thus,

(3.33)
But since (Eq. 3.27), the overall limits obtained are:

(3.34)
The second integral is limited to the exact point:

(3.35)
due to constraint (3.29).

4
Here, it was considered that ∫ , since the integral is an infinite sum of differentials,

and in this case we are considering a single case when . Thus, the result of the integral (or
infinite sum) simply becomes a single differential .

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (20 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Then, the resulting constrained distribution for will be (from Eq. 2.64, 3.31 and 3.34):

∫ ∫
( )
( )

(3.36)
Eq. (3.36) represents a uniform distribution of values in the range .
On the other hand, Eq. (3.31) can be alternatively determined as follows:

( )
∫ ∫ ∫ ∫
( )

(3.37)
and therefore:

∫ ∫
( )
( )

(3.38)

also representing a uniform distribution of values, but in this case in the range .

Figure 6 illustrates the probability density functions of both variables before and after imposing
the constraint.

Figure 6. Original (solid blue line) and constrained (dashed red line) probability density function of
variables (left plot) and (right plot) for Example # 6.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (21 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

3.7. Example # 7: Multivariate Equality Constraint imposed on two Normal Variables

Now consider two independent standard normal random variables and ,

subject to the same constraint given in the previous example (Eq. 3.29). While the membership
function remains unchanged (Eq. 3.30), the probability of the constraint set now becomes:

( )
∫ ∫
( ) √ √
( )
( )
∫ ∫

√ √
(3.39)

Figure 7. Original (solid blue line) and constrained (dashed red line) probability density functions of
variables and for Example # 7.

The corresponding constrained probability density functions will be:

( )
∫ ( )
( )
√ √
( ) √
√ √
(3.40)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (22 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( )
∫ ( )
( )
√ √
( ) √
√ √
(3.41)

Both expressions correspond to normal distributions with a mean value of and standard
deviation √ . Figure 7 illustrates the probability density functions before and after imposing
the constraints. In this example, both variables behave identically, but this is not necessarily
always the case. Notice also that the constrained variables are no longer independent from
each other.

3.8. Example # 8: Multivariate Inequality Constraint imposed on two Normal Variables

On the same pair of independent normal random variables and , the

following inequality constraint is imposed:

(3.42)
with membership function:

( ) ( ) ( )
{

(3.43)
where represents Heaviside’s step function, defined as ( ) { .

The probability of the constraint set can be determined as follows5:

( )
∫ ∫ ( ( ) ( )
)
√ √
√
∫ (∫ )
√

∫ (√ )
√
(3.44)

5
A numerical evaluation of the last integral is required. The value obtained for this integral can be found
here: integral(exp(-x^2/2)*erf(sqrt((1-x^2)/2))), from -1 to 1 - Wolfram|Alpha (wolframalpha.com).

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (23 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Now, the constrained probability density functions will be:

∫ ( ( ) ( )
)
√ √
( )
( )

√
∫
√
(√ )

(3.45)

(√ )
( )

(3.46)
The constrained probability density function obtained is compared to the original
unconstrained probability density function in Figure 8.

Figure 8. Original (solid blue line) and constrained (dashed red line) probability density functions of
variables and for Example # 8.

3.9. Example # 9: Multivariate Inequality Constraints imposed on Three Variables: Normal,

Uniform and Exponential

Increasing the level of complexity, in this example three different standard random variables
are considered: , , and , with probability densities given by Eq. (3.3),
(3.9), and (3.47), respectively.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (24 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( ) ( )
{

(3.47)
The constraints imposed on these variables are the following:

(3.48)
with membership function:

( ) ( ) ( )
{

(3.49)
The probability of the constraint set is for this problem:

( )
∫ ∫ ∫ ( ( ) ( )
)

∫ ∫ ∫
√

∫ (∫ (∫ ) )
√

∫ (∫ ∫ )
√

( ∫ ∫ )
√

( ) ( ) √
√ √
(3.50)
And the resulting constrained probability density functions are:

∫ ∫
√ ( )
( )
( )

(3.51)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (25 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

√
∫ ∫
√ √
( )
( )

( (√ ) (√ ))

(3.52)
√
∫ ∫
√ √
( )
( )

√ (√ )

(3.53)
Figure 9 shows the probability density functions of the original and constrained variables.

Figure 9. Original (solid blue line) and constrained (dashed red line) probability density function of
variables (top left plot), (top right plot) and (bottom plot) for Example # 9.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (26 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

3.10. Example # 10: Multivariate Constraint considering Probability Functions

Let us consider four independent discrete uniform binomial variables , , and , such
that:
{ }
( )
{ { }
{ }
(3.54)

These variables are subject to the following constraints:

( ) ( )
( ) ( )
( ) ( )
( ) ( )
(3.55)
Notice that this problem is equivalent to a magic square [7] having four cells. Also notice here
the difference between the probability of the original randomistic variable, like for example
( ) which is equal to ( ) , and the probability of the constrained variable, like
for example ( ) ( )
which is when and when .

In fact, the probability distribution of the original variables would satisfy all constraints if they
were referred to the probabilities of the randomistic variable, which is not the case.

In this problem, we want to find the four values , , , and that satisfy the four
constraints to . However, one of the constraints is redundant, as it can be obtained from
the previous three expressions:

(3.56)
So, actually there are only three equality constraints leaving one degree of freedom available.
Unless we assume the value of one of the variables (the available degree of freedom) no
specific solution would be obtained.

If we assume ( ) ( )
, then we obtain from Eq. (3.55):

( ) ( )

( ) ( )

( ) ( )
(3.57)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (27 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

Since the original variable is binomial, then:

( ) ( )

( ) ( )
(3.58)
Alternatively, if we assume ( ) ( )
, then we obtain:

( ) ( )

( ) ( )

( ) ( )

( ) ( )
(3.59)
These two solution sets can be represented by arranging the four variables in a table, as
depicted in Figure 10.
2x2 Table Solution 1 Solution 2

1 2 2 1
2 1 1 2

Figure 10. Left: arrangement of the four variables considered in Example # 10. Middle: Solution #1.
Right: Solution #2.

These solutions can also be found algebraically as follows.

Solution #1:
( )

( )
[ ] [ ] [ ]
( )

[ ( )]
(3.60)
Solution #1:
( )

( )
[ ] [ ] [ ]
( )

[ ( )]
(3.61)

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (28 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

4. Summary

A randomistic variable can be constrained so that only a determined set ( ) of realization

values is feasible. The feasible set can be represented in multiple possible ways, including
equalities, inequalities, but in general, they are described by Boolean membership functions
( ( ) ), such that:

( ) {
(2.1)

This restriction imposed on the variable may result in a different distribution of probabilities for
the values in the constrained variable ( ( ) ), compared to the probability distribution in the
original variable ( ( ) ):

( )
( ) ( )
( )
(2.4)
where

( ) ∑ ( ) ( )

(2.2)
In the case of continuous variables, the probability density function of the constrained variable
will be:
( )
( ) ( )
( )
(2.5)
where

( ) ∫ ( ) ( )

(2.3)
If the restrictions considered involve two or more ( ) independent randomistic variables
simultaneously, then the constrained probability function of the arbitrary variable
becomes:
∑ ∑ ∏
( ) ( )

( )
( )

(2.63)
with

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (29 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

( )
∑ ∑ ∑ ∏
( ) ( )

(2.61)
And for continuous variables:

∫ ∫ ( ) ( )
∏ ( ( )
)
( )
( )

(2.64)
with

( )
∫ ∫ ∫ ( )
∏( ( )
)

(2.62)
The expressions for the constrained randomistic variable are valid as long as the probability of
the constraint set ( ) is greater than zero. If the ( ) is exactly zero, then an empty
constraint set is obtained, and no solution exists.

Acknowledgment and Disclaimer

This report provides data, information and conclusions obtained by the author(s) as a result of original
scientific research, based on the best scientific knowledge available to the author(s). The main purpose
of this publication is the open sharing of scientific knowledge. Any mistake, omission, error or inaccuracy
published, if any, is completely unintentional.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-
for-profit sectors.

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC
4.0). Anyone is free to share (copy and redistribute the material in any medium or format) or adapt
(remix, transform, and build upon the material) this work under the following terms:
 Attribution: Appropriate credit must be given, providing a link to the license, and indicating if
changes were made. This can be done in any reasonable manner, but not in any way that
suggests endorsement by the licensor.
 NonCommercial: This material may not be used for commercial purposes.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (30 / 31)
 Constrained Randomistic Variables
Hugo Hernandez
ForsChem Research
[email protected]

References

[1] Hernandez, H. (2022). Standard Deterministic, Standard Random, and Randomistic Variables.
ForsChem Research Reports, 7, 2022-06, 1 - 18. doi: 10.13140/RG.2.2.36316.87688.
[2] Hernandez, H. (2020). On the Discreteness of Measured Variables and the Continuous
Approximation. ForsChem Research Reports, 5, 2020-20, 1-18. doi: 10.13140/RG.2.2.27740.00646.
[3] Hernandez, H. (2017). Multivariate Probability Theory: Determination of Probability Density
Functions. ForsChem Research Reports, 2, 2017-13, 1-13. doi: 10.13140/RG.2.2.28214.60481.
[4] Hernandez, H. (2024). Calculus of Probabilities of Randomistic Variables. ForsChem Research
Reports, 9, 2024-01, 1 - 31. doi: 10.13140/RG.2.2.27248.23044.
[5] Hernandez, H. (2018). Introduction to Randomistic Optimization. ForsChem Research Reports, 3,
2018-11, 1-25. doi: 10.13140/RG.2.2.30110.18246.
[6] Hernandez, H. (2023). The Periodic Equality Operator ( ). ForsChem Research Reports, 8, 2023-04,
1 - 8. doi: 10.13140/RG.2.2.36167.57768.
[7] Euler, L. (1862). De quadratis magicis. Opera Posthuma Mathematica et Physica. Academiae
Scientiarum Petropolitanae, Petropoli. pp. 140-151. https://scholarlycommons.pacific.edu/euler-
works/795. English translation: arxiv.org/pdf/math/0408230.pdf. French translation:
arxiv.org/pdf/math/0607468.pdf.

07/03/2024 ForsChem Research Reports Vol. 9, 2024-03

10.13140/RG.2.2.12411.73764 www.forschem.org / t.me/forschem (31 / 31)