Chapter 10 – Introduction to Simulation Modeling
19. A correlation matrix must always be symmetric, so that the correlations above the diagonal are a mirror image of those
below it.
20. Correlation between two random input variables may change the mean of an output, but it will not affect the
variability and shape of an output distribution.
Exhibit 10-1
A company is in the planning phase of constructing a new production facility. It wants to build a simulation model for the
economics of the facility, and one key uncertain input is the construction cost. For each of the scenarios in the questions
below, choose an “appropriate” distribution, together with its parameters, and explain your choice.
21. Refer to Exhibit 10–1. Company management currently has no idea what the distribution of the construction cost is.
All they can state is that “we think it will be somewhere between $5,000,000 and $8,000,000.”
The “no idea” suggests the uniform distribution, with a lower bound of $5M and an upper bound of $8M.
22. Refer to Exhibit 10–1. A little later on, management still believes the upper and lower bounds for the costs are $5M
and $8M, but now they can also state that “we believe the most likely value is about $6.5M.”
This suggests a triangular distribution, with a min of $5M, most likely value of $6.5M, and max of $8M.
23. Refer to Exhibit 10–1. Management believes the facility construction time will be somewhere from 5 to 9 months.
They believe the probabilities of the extremes (5 and 9 months) are both 10%, and the probabilities will vary linearly from
those endpoints to a most likely value at 7 months.
This is a general discrete distribution. We just have to choose the probabilities of the values 5 to 9 so that they
increase and then decrease linearly, and add up to 1: P(5)=0.1, P(6)=0.225, P(7)=0.35, P(8)=0.225, P(9)=0.1.
24. Refer to Exhibit 10–1. Engineering also believes the construction time will be from 5 to 9 months. However, they
believe that 7 months is twice as likely as either 6 months or 8 months and that either of these latter possibilities is three
times as likely as either 5 months or 9 months.
This is another general discrete distribution, where we have to choose the probabilities so that they have the
specified ratios, and add up to 1: P(5)=0.072, P(6)=0.215, P(7)=0.43, P(8)=0.215, P(9)=0.072.
25. If you add n lognormally distributed random numbers, the mean of the distribution for the sum is the sum of the
individual means, and the variance of the distribution of the sum is the individual variances. This result is difficult to
prove mathematically, but it is easy to demonstrate with simulation. To do so, run a simulation where you add three
1