Module 1 reflection


Lecture

2023-09-29

Reflections

Today

  1. Reflections

  2. Revisions

  3. Practice problems

Exam 1

  • Average: 47.5
  • Standard deviation: 15.9
  • Maximum: 70
  • Median: 49.5
  • Easiest: maximum likelihood
  • Hardest: computation, conditional probability

How I’ll support your learning

  • Reorganized schedule – two modules not three
  • Practice problems at start of class
  • More links to suggested reading for each lecture

How you need to support your learning

  • Come to office hours
  • Review lecture notes
    • Don’t just read
    • Work through examples
    • Understand why
    • Ask questions

I have encouraged you to use tools (Copilot, GPT, etc) to help you with computing syntax so you can spend more time on understanding concepts.

Revisions

Today

  1. Reflections

  2. Revisions

  3. Practice problems

Revisions

  • Exam I is graded
  • Final grades will be curved, not each exam
  • Revisions will be allowed

Due Date

  • Friday, October 6th at 11:00AM.
  • Hand in to me in class
    • Legible and clear handwritten work OR
    • Type up your work1
    • Hand in your original exam with your revisions

Academic integrity

  • Subject to Rice Honor Code
  • DO: consult class notes, textbooks, linked resources, or write your own code
  • DO NOT: Consult with a classmate, search the internet, usie AI chat tools, or otherwise collaborate is not permitted
  • Ask if you’re not sure

If you have questions about what is permitted, please ask.

Instructions

For each problem:

  1. State how many points you earned on the original exam.
  2. Derive the correct answer. Your answer should be clearly written and easy to follow.
  3. Explain why your original answer was incorrect and what confused you (no revisions are needed if your original answer was correct!)

Grading

  1. On each problem, you will earn up to 60% of the points you missed
    1. Grading on the revision will be more strict than on the original exam
  2. Additional 10% for clear and insightful explanation of your original mistake (where applicable)
  3. T/F questions: up to 50% for a clear and correct explanation of why the statement is true or false
  4. If your revision is worse than the original, your score will not be lowered.

Suggestions

  • Precipitation:
    • CDF: \(F(x) = P(X \leq x)\)
    • Can assume that \(F(0)=0.6\)
  • Conditional probability: think hard about how to define \(A\) and \(B\)
  • Return period: read the wording carefully (“according to the distribution shown…”, “…true return period”)
  • Recording of review session

Practice problems

Today

  1. Reflections

  2. Revisions

  3. Practice problems

Bayes Rule

A doctor is called to see a sick child. The doctor has prior information that 90% of sick children in that neighborhood have the flu, while the other 10% are sick with measles. Assume for simplicity that there are no other illnesses and that no children have both the flu and measles.

A well-known symptom of measles is a rash. The probability of having a rash if one has measles is 0.95. However, occasionally children with flu also develop rash, and the probability of having a rash if one has flu is 0.08. Upon examining the child, the doctor finds a rash. What is the probability that the child has measles?

Analytic posterior

We collect some count data and model it using a Poisson likelihood. The Poisson likelihood is given by: \[ p(y_i | \lambda) = \frac{\lambda^{y_i} e^{-\lambda}}{y_i!} \] where \(y_i\) is the number of counts and \(\lambda\) is the rate parameter of the Poisson distribution. We want to do inference on \(\lambda\). We have a prior belief that \(\lambda\) is distributed as an Exponential distribution with the modified parameterization: \[ p(\lambda | \theta) = \frac{1}{\theta} e^{-\frac{\lambda}{\theta}} \] After collecting data \(y_1, y_2, \ldots, y_n\), what is the posterior distribution of \(\lambda\) given our prior parameter \(\theta\)?## Next week:

  • Monday: Gridded climate data lab
  • Wednesday: Generalized Linear Models
  • Friday: Loss functions