Introduction


Plagiarism and collusion

As this is a group project, collusion and academic misconduct cases will apply to all members in the group. All group members bear full responsibility. If a student suspects any potential misconduct cases from other groupmates, they should inform me immediately.

Please familarize yourself with the following guidance on academic integrity

In addition please note that parts of your submission may be scanned using similarity detection software. If any breach of the assessment regulations is suspected, it will be investigated in accordance with UCL’s Student Academic Misconduct Procedure.

By ticking the submission declaration box in Moodle you are agreeing to the following declaration:


Declaration: I am aware of the UCL Statistical Science Department’s regulations on plagiarism for assessed coursework. I have read the guidelines in the student handbook and understand what constitutes plagiarism. I hereby affirm that the work we are submitting for this in-course assessment is entirely our own.

Anonymous Marking

Please do not write your names anywhere on the submission. Please include only your student numbers as a proxy identifier.


Problems

Question 1 [20 points]

Consider \(n=120\) independent \(M/M/1\) queues. Items arrive to each queue at rate \(\lambda = 5\). Each queue has a server who serves items at possible different rates \(\mu_1, \ldots, \mu_{120}\). The rate of the server is given by a linear model,

\[\mu_i= mx_i + b + \epsilon_i,\]

where \(\epsilon_i\) are iid normal noise variables with mean zero and \(x_i\) are the initial appraisals of the servers, given by a real number in \([5,10]\). The appraisals are given by a text file here. We also have daily \(k=50\) inspections, where we record the number of items in each queue. These numbers are given by a text file here.

  • Find reasonable estimators for \(m\) and \(b\). Explain and justify how you arrive at your estimates and state any assumptions that are necessary for your inference. [10 points]
    • Apply your method on a data set that you generated, so that you know that it works. [5 points]
    • Apply your method on the given data set to find \(m\) and \(b\). [5 points]

Question 2 [15 points]

Consider the renewal process \(N\) with inter-arrival times that are independent and uniform over \([0.5, 1.5]\). For all integers \(k\) consider the following limit:

\[\lim_{t \to \infty}\mathbb{P}[N(t+1) - N(t)=k].\]

  • For what values of \(k\) is limit non-zero. [2 points]
  • Use simulations to estimate these probabilities. [3 points]
  • Compute the limit, analytically, pen and paper style. [10 points]

Question 3 [10 points]

Consider the following scenario. You know from your running apps that you can run \(1\) mile pretty reliably, meaning \(99\) percent of the time, you can run a mile between \(9\) and \(10\) minutes. A \(M(5)/M(5.1)/1\) queue is \(1\) mile away–here it is a rate of \(5\) customers per minutes. Estimate the probability that that you will make to through the queue within \(20\) minutes. Make clear any assumptions you are using for your calculations/simulations. Part of this exericse is to come up with reasonable modelling assumptions. Give one answer than you can do without any complicated calculations–like one that you can perform while you are running and deciding if you will make it or now, and give another answer that you think is more accurate and makes better use of the available information. Discuss the differences in your numerical answers.

Question 4 [10 points]

Consider two queues both with exponential service rate \(\mu\), where \(\mu > \lambda>0\)—for both queues the arrival rate is \(\lambda\). Suppose also that I know more about the distribution of the arrivals for the queues. I am interested in picking the one with the least average waiting time in the system, where both queues have been in operation for a long time. Does it matter which queue I pick? Support your answer with code and/or theory.

Question 5 [10 points]

Consider a Poisson point process \(\Gamma\) of intensity \(2\) on the interval \([-1,1]\). Suppose that I modify this point process on the interval \([0,1]\), where I delete points independently with probability \(\tfrac{1}{2}\), and we obtain a new point process \(\Pi\) consisting of possibly fewer points than \(\Gamma\).

  • Let \(N = \Pi[-1,1]\) be the total number of points of \(\Pi\).
    • What is \(\mathbb{E} N\)? [2 points]
    • What is the distribution of \(N\)? [3 points]
  • Suppose that there is exactly one point in the interval \([-0.5,0.5]\)—let \(V\) be the location of this point. What is the distribution of \(V\)? [5 points]



Endnotes

  • Source
  • Version: 13 December 2024