Simulation Simulation helps us to understand the consequences of uncertainty. The probability of completion time or cost of a single task, a set of tasks, or the project does not exceed a specific value. We can estimate Beta distribution with Triangular distribution. Traditional Statistics vs. Simulation Both must enumerate all the paths Traditional statistics assume path interdependence while simulation does not 3/23/2016 Ardavan Asef-Vaziri 5-3-1 Triangular Probability Distributions a m b a = smallest value the variable can assume b = largest value the variable can assume m = most likely value the variable can assume Expected Value of x Variance of x 3/23/2016 µ = (a +m+ b)/3 s 2 = (a2 + m2 +b2 – ab- am-bm)/18 Ardavan Asef-Vaziri 5-3-2 Generate a Triangular Random Variable Compute the height of this triangle AreaL (m a) 1 ma (b a) b a Similarly, the area to the right is H= 2/(b-a) a m b 1 bm AreaR (b m) (b a) b a Generate a random number if rand() ≤ (m-a)/(b-a) we are on the left, otherwise on the right 3/23/2016 Ardavan Asef-Vaziri 5-3-3 Triangular Probability Distributions Suppose we are on the left Then the triangular random variable is something like X H h a X m b What is the height associated with X? h X a H ma hH X a ma The area to the left of X is then 3/23/2016 Ardavan Asef-Vaziri 2( X a) h (m a)(b a) Area LX h( X a ) 2 5-3-4 Triangular Probability Distributions h( X a ) Area LX 2 2( X a) h (m a)(b a) H h a ( X a) (m a)(b a) X m b 2 AreaLX The area to the left of X was our rand() which happened to drop on the left 3/23/2016 ( X a) 2 Rand () (m a)(b a) Ardavan Asef-Vaziri 5-3-5 Triangular Probability Distributions ( X a) Rand () (m a)(b a) 2 H h a ( X a) (m a)(b a) Rand () X m b 2 X a (m a)(b a) Rand () Generate a random number if ≤ (m-a)/(b-a) we are on the left, and implement the above equation to generate the random variable, otherwise we are on the right 3/23/2016 Ardavan Asef-Vaziri 5-3-6 Triangular Probability Distributions We showed that for X on the left AreaLX ( X a) 2 (m a)(b a) H h a m b X Similarly, we can show that if X is on the right of m, the area to the right of h is (b X ) (b m)(b a) 2 AreaRX This is the area to the right of X, but the area to the left of X was equal to Rand() The area to the right of X is equal to 1- Rand() 3/23/2016 Ardavan Asef-Vaziri 5-3-7 Triangular Probability Distributions (b X ) (b m)(b a) 2 AreaRX H (b X ) 1 Rand () (b m)(b a) h 2 a m b X (b X ) 2 (b m)(b a)(1 Rand ()) X b (b m)(b a)(1 Rand ()) 3/23/2016 Ardavan Asef-Vaziri 5-3-8 Triangular Probability Distributions Generate a random number If ma Rand () ba we are on the left, otherwise on the right. If on the Left X a (m a)(b a) Rand () If on the Right X b (b m)(b a)(1 Rand ()) 3/23/2016 Ardavan Asef-Vaziri 5-3-9 Triangular Probability Distributions Suppose a=5, b= 20, m=10 ma 10 5 AreaL 0.333333 ba 20 5 20 10 bm 0.666667 AreaR 20 5 ba (m a )(b a) (10 5)( 20 5) 75 (b m)(b a ) (20 10)( 20 5) 150 3/23/2016 Ardavan Asef-Vaziri 5-3-10 Triangular Probability Distributions Generate Rand () If Rand () 0.33333 Otherwise 3/23/2016 X 5 75Rand () X 20 150(1 Rand ()) Ardavan Asef-Vaziri 5-3-11 Generate 3/23/2016 Ardavan Asef-Vaziri 5-3-12 Histogram 3/23/2016 Ardavan Asef-Vaziri 5-3-13 Descriptive Statistics Column1 Mean Standard Error Median Mode Standard Deviation Sample Variance Kurtosis Skewness Range Minimum Maximum Sum Count 3/23/2016 11.70 0.102 11.42 #N/A 3.22 10.37 -0.61 0.30 14.30 5.22 19.51 11701 1000 Ardavan Asef-Vaziri 5-3-14 Assignment Problems: Ch5 Problem 29(it refers to Problem 27 not 26). All activities have beta distribution. Solve the problem using (a) Normal distribution for all activities, (b) Triangular distribution for all activities If you solve it using excel, you do not need to solve it using CB. If you solve it using CB, also solve it using excel. You will see solving it simply using excel is not more difficult than CB. 3/23/2016 Ardavan Asef-Vaziri 5-3-15