Вы находитесь на странице: 1из 5

The Triangular Probability Distribution

The triangular distribution is one of the most widely used distributions in


simulation analysis. It is often the distribution of choice when little is known
about the probability distribution. It has three parameters:
• the minimum
• the maximum
• the mode
From these, we can define the nature and characteristics of the entire
distribution.

The triangular distribution is comprised of either one or two triangles, as shown


below:

Min = Max Min Mode Max Min Max =


Mode Mode

Recalling that any probability density function must sum to 1, and that the
formula for the area of a triangle is area = (½)(base)(height) = (½)(B)(H), we can
easily determine the height of the triangular distribution:

1 = (½)(B)(H)
H = 2/B = 2/(Max – Min)

In order to use this as a probability distribution, we need to be able to determine


the area corresponding to certain values of the distribution and, conversely, the
values of the distribution corresponding to specific areas of under the curve. As
the sides of congruent triangles are always proportional, the area of a portion of
the distribution can be determined by finding the lengths of the sides of the
corresponding relevant triangles.

Let’s start with the left side of the


distribution. Suppose we have a
point, q, and wish to know the
cumulative probability to that value,
shown by the shaded are in the graph
at the right. We know that the base of
the shaded triangle is (q – Min). The
height, Hq will be determined by Min q Mode Max

The Triangular Probability Distribution Page 1


scaling the height of the triangle that forms the left half of the distribution:

Hq H
=
(q − Min) (Mode − Min)
2 (Max − Min)
=
(Mode − Min)
2
=
(Mode − Min)(Max − Min)
2(q − Min)
Hq =
(Mode − Min)(Max − Min)
Therefore, the area of the smaller, shaded triangle, which corresponds to the
cumulative probability associated with the value q is:

2(q − Min) 1
P( x < q) = × (q − Min) ×
(Mode − Min)(Max − Min) 2

=
(q − Min)
2

(Mode − Min)(Max − Min)


To determine a cumulative
probability above the mode, on the
right side of the distribution, we work
with the complementary probability
as shown at the right:

Min Mode q Max

P( x < q) = 1 − P( x > q)
2(Max − q) 1
= 1− × (Max − q) ×
(Max − Mode )(Max − Min) 2

= 1−
(Max − q)2
(Max − Mode )(Max − Min)

Numerical Example

Suppose we have a triangular distribution with a minimum value of 4, a mode of


6 and a maximum value of 12. What are the cumulative probabilities of 5 and 10,
respectively?

The Triangular Probability Distribution Page 2


P( x < 5) =
(q − Min)2
(Mode − Min)(Max − Min)
=
(5 − 4)2 = 1 = 1
(6 − 4)(12 − 4) (2)(8) 16
P( x < 10) = 1 − P( x > 10)

= 1−
(12 − 10 )
2

(12 − 6)(12 − 4)
22 4 11
= 1− = 1− =
(6)(8) 48 12

Using the Triangular Distribution in Simulations

The classic method to simulate using any probability distribution is to generate a


random number between 0 and 1 and use this value as the cumulative probability
to generate a random variate. Using the two formulas that we just used in the
examples above, the cumulative probability will now be known, and the value of q
will be our unknown.

First, because the computations differ on the different sides of the distribution, we
must determine in which side of the distribution the random number will fall. The
areas of the sides of the distribution are proportional to their respective bases, so
that if RN < (Mode – Min)/(Max – Min), it corresponds to the left side of the
distribution; otherwise it is in the right side.

Solving for the values of the random variates, first on the left side:

P( x < q) = RN =
(q − Min)
2

(Mode − Min)(Max − Min)


(q − Min)2 = (RN)(Mode − Min)(Max − Min)
(q − Min) = (RN)(Mode − Min)(Max − Min)
q = Min + (RN)(Mode − Min)(Max − Min)

On the right side:

The Triangular Probability Distribution Page 3


P( x < q) = 1 − P( x > q) = 1 − RN

= 1−
(Max − q)2
(Max − Mode )(Max − Min)
(Max − q) = (1 − RN)(Max − Mode )(Max − Min)
2

(Max − q) = (1 − RN)(Max − Mode )(Max − Min)


q = Max − (1 − RN)(Max − Mode )(Max − Min)

Excel™ Considerations

With simulation software, such as Arena and Crystal Ball, the triangular
distribution is one of the most widely used distributions. While many distributions
are built into Excel™, the triangular distribution is not one of them. However,
implementing a triangular distribution formula is quite straightforward. We cannot
have the random number generation, RAND(), embedded into our formula as we
need to refer to the random number generated more than once, which is not
possible in Excel™. Therefore, we will generate the random number externally,
and reference it within the formula.

The formula would use the IF() function to determine which side corresponds to
the random number generated. The structure of the IF function is simple:

=IF(expression, what is returned if true, what is returned if false)

For our triangular distribution, the formula first determines which side of the
distribution corresponds to the random number and then evaluates the
appropriate formula.

=IF(RN < (Mode – Min)/(Max – Min),LS formula, RS formula)


where
LS formula is Min + (RN)(Mode − Min)(Max − Min)
and
RS formula is Max − (1 − RN)(Max − Mode )(Max − Min)
The picture below shows an Excel™ implementation of a triangular distribution:

The Triangular Probability Distribution Page 4


A B C D
1 Triangular Distribution
2
3 Min 4 Parameter Value
4 Max 12 Parameter Value
5 Mode 6 Parameter Value
6 Range 8 Range = Max - Min
7 LSR 2 LSR = Left Side Range = Mode - Min
8 RSR 6 RSR = Right Side Range = Max - Mode
9
10 RN Variate
11 0.606164 7.65211 =IF(A12<($B$7/$B$6),
12 0.80169 8.91473 $B$3+SQRT(A12*$B$7*$B$6),
13 0.704062 8.231043 $B$4-SQRT((1-A12)*$B$8*$B$6))
14 0.722482 8.350222

The Triangular Probability Distribution Page 5

Вам также может понравиться