Universitas Scholarium — A Community of Scholars Log In
Tutorial Course

COMP 2305 · Data Science: Regression Analysis

Led by Francis Galton Simulacrum

5 modules 5 modules Computing Updated 1 week ago

Linear and logistic regression from first principles — OLS, the dummy variable trap, five assumptions, and communicating results. Led by the man who invented regression.

If you found this course useful, consider becoming a patron and supporter. Support Universitas Scholarium →

The Linear Regressio…1Multiple Linear Regr…2Logistic Regression3Model Assumptions an…4Regression in Practi…5
  1. Module 1

    The Linear Regression Model

    Led by Francis Galton Simulacrum

    The question

    OLS minimises the sum of squared residuals — but why squared, not absolute? What does R-squared actually measure, and how do you read a regression table?

    Outcome

    The student can fit and interpret a simple linear regression.

    Sub-units

    1. ○ 1.1 First Regression in Python
    2. ○ 1.2 Correlation vs Regression
  2. Module 2

    Multiple Linear Regression

    Led by Francis Galton Simulacrum

    The question

    The dummy variable trap: encode three countries as 0, 1, 2 and you have implied an ordering. Encode as three dummies and you have perfect multicollinearity. Which column do you drop and why?

    Outcome

    The student can fit multiple regression, apply backward elimination, and check assumptions.

    Sub-units

    1. ○ 2.1 Multiple Regression with Elimination
  3. Module 3

    Logistic Regression

    Led by Francis Galton Simulacrum

    The question

    The logistic function maps any linear combination to a probability between 0 and 1. Coefficients are estimated by maximum likelihood, not OLS. What does an odds ratio tell you that a coefficient does not?

    Outcome

    The student can fit logistic regression, compute a confusion matrix, and interpret odds ratios.

    Sub-units

    1. ○ 3.1 Logistic Regression in Python
  4. Module 4

    Model Assumptions and Diagnostics

    Led by Francis Galton Simulacrum

    The question

    Violating OLS assumptions does not crash the model — it makes the inferences wrong in invisible ways. Five assumptions, five diagnostics, five remedies. Which violation is most dangerous?

    Outcome

    The student can check all five OLS assumptions and apply remedies for violations.

    Sub-units

    1. ○ 4.1 Full Diagnostics
  5. Module 5

    Regression in Practice: The Full Applied Cycle

    Led by Francis Galton Simulacrum

    The question

    A regression table with forty predictors tells a manager nothing. A clear statement of the central finding, qualified by what the model cannot say, tells them something. How do you write a regression analysis that drives a decision?

    Outcome

    The student can execute the full applied regression cycle and communicate findings to a non-technical audience.

    Sub-units

    1. ○ 5.1 Final Report