Here are (Part I) T or F Questions and (Part II) Multiple choice questions:

1) If random variables A and B are unconditionally independent, then A) both A and B are causes of another variable C. B) E[A|B] = E[A]. C) If A and B are both correlated with a variable C, then estimating the causal effect of A on B will always suffer omitted variable bias. D) A and B are commonly influenced by some other variables. E) None of the above. 2) When a constraint line bounding a feasible region has the same slope as an iso- profit line, A) the problem involves redundancy. B) an error has been made in the problem formulation. C) a condition of infeasibility exists. D) there may be more than one optimum solution. E) None of the above. 3) We want to ship 100 different items and we have 10 trucks with enough capacity. We want to minimize the total wasted capacity. How many decision variables do we need in our model? A) 100 B) 110 C) 1000 D) 1010 E) None of the above 4) A company uses a continuous review system to manage inventory, (0:100, r=20). The manager just checked their inventory and found out the current inventory in stock is 18. The manager should A) Place an order of 80 units B) Place an order of 100 units C) Do not place an order D) Place an order of 82 units E) None of the above. Part I (True or False): 1) In linear programming, if the optimal solution changes when we increase the amount of a resource, its shadow price may not change. 2) When building a Monte Carlo simulation model of a single-server, single- queue sen-ice system, if we know that customer arrivals follow Poisson distribution and the service time follows uniform distribution, then we cannot simulate the arrival time or departure time of a customer. 3) In a mixed integer program for assigning 10 different jobs to 10 workers, if every worker must work on a job, it is possible to introduce only 9 binary variables in total. 4) If two consulting rms independently investigate the market prospect of a product, then their predictions of the market condition will be independent. 5) Suppose we randomly selected a group of students and divided them into treatment and control groups according to their personal computer (PC) ownership. Students in the treatment group have PCs and those in the control group do not. We can use the difference between the average GPA of treatment group and that of the control group to estimate the causal effect of PC on GPA. 6) In decision making under risk, we can derive the whole payoff table from the opportunity loss table alone. 7) If A and B are independent when we condition on C, then A can be an instrumental variable to estimate the impact of C on B. 8) If we want to estimate the impact of decision D on output Y, we must make sure the decision is completely random in the historical data. The following three questions are based on the causal graph shown below. 0 5) If we want to estimate how 0 influences Y on aggregate level, using conditioning strategy, we can condition on A) D and N B) B and N C) D and M D) B and D E) None of the above. 6) Assume the causal influences are all linear in this system. Which of the following models can correctly estimate the aggregate effect of D on Y? A) Y: a + [3D + s B)Y=a+[3D+yO+BB+ C)Y=a+D+g/O+5M+e D)Y=a+D+yO+BB+5M+e E) None of the above. 7) If the model isY=a+D+yO+5M+e, how should we interpret the estimated 6'? A) The expected change of Y when we change D by one unit B) The aggregate effect of M and some other factors on Y C) The expected change of Y when we change M by one unit while other factors remain unchanged D) The additional expected value of Y for the group of individuals with M = 1 compared against the group with M = 0. E) None of the above