Prediction thinking
Students learn to define a prediction question before choosing an algorithm.
Module 1
Foundations of Machine Learning in Biostatistics
This module builds the core language of medical machine learning: prediction, explanation, causal caution, learning types, validation, overfitting, leakage and responsible biostatistical reporting.
Module aim
Build judgement before algorithms.
The purpose of this module is to help students understand what a medical prediction model can and cannot support before moving into supervised learning methods.
5
Lessons
R
Coding labs
Complete
Module status
Clinical
Prediction focus
Interpretation discipline
The module separates prediction, explanation and causation to avoid unsafe claims.
Validation awareness
Training, testing, overfitting, generalisation and leakage are treated as core ideas.
Module lessons
Study the lessons in order.
Each lesson adds one layer of judgement: what ML means in health data, how to avoid causal overclaiming, how to classify learning tasks, how to validate honestly and how to report a complete biostatistical ML workflow.
1.1
Lesson 1.1
What is machine learning in biostatistics?
Understand machine learning as a prediction-focused workflow for health data, clinical questions and biostatistical decision-making.
Open lesson →
1.2
Lesson 1.2
Prediction vs explanation vs causal thinking
Learn why prediction, explanation and causation are different aims, and why reports must not overclaim what a model supports.
Open lesson →
1.3
Lesson 1.3
Types of learning
Classify supervised, unsupervised and semi-supervised learning problems using the outcome structure and modelling goal.
Open lesson →
1.4
Lesson 1.4
Training, testing, overfitting and generalisation
Understand train/test validation, overfitting, generalisation, unseen patients and the danger of data leakage.
Open lesson →
1.5
Lesson 1.5
Biostatistical ML workflow
Bring the module together through a careful workflow: clinical question, predictors, validation, threshold judgement and reporting.
Open lesson →
Learning route
Finish this module before moving into supervised learning.
Module 2 assumes that students understand prediction questions, predictor timing, validation logic, overfitting, leakage and the difference between prediction, explanation and causation.
Continue to Module 2 →Case study route
Apply Module 1 ideas to diabetes risk prediction.
After completing the foundation lessons, use the case study to see how prediction, validation, thresholds and reporting appear in a health-data example.
Open case study →