AI Creators
StatQuest with Josh Starmer
StatQuest: Random Forests Part 2: Missing data and clustering
Jun 8, 2026
Key Points
- 1.Random forests can handle missing data through iterative refining of guesses.
- 2.Missing values are initially estimated and gradually improved by calculating sample similarities.
- 3.The proximity matrix tracks similarities between samples to enhance predictions.
Summary
Understanding Missing Data in Random Forests
Random forests address missing data in two contexts: data missing from the training set and data missing from new samples. Initial estimates for missing values can be made based on common values or median statistics from similar samples.
Using Proximity Matrices to Determine Sample Similarity
Sample similarity is established through a proximity matrix, which records how often samples end up in the same leaf node across different trees. The proximity values are calculated by running data through the random forest, allowing for better estimates of missing data.
Iterative Process for Refining Missing Values
To improve initial estimates, the process involves building the random forest multiple times, updating the proximity matrix after each iteration. The goal is to converge on accurate predictions of missing values by calculating weighted averages based on sample proximities.
From Proximity to Distance Matrices
After calculating the proximity values, one can derive a distance matrix for visual representation. The values in the matrix can be used to create heat maps or MDS plots, illustrating the relationships between samples.
Worth watching for
This video is intended for data scientists and statisticians interested in advanced techniques for handling missing data with random forests.