A high can make a regression model look impressively accurate — but this number can be deceptive. If you want to understand why a high is not always a sign of a good model, read on! In the post, Learning Data Science: Modelling Basics, we built a simple model to predict income from age. R … Continue reading "Learning Data Science: Why a High R^2 Can Be Misleading"

R-bloggers

R² (coefficient of determination) is a commonly used metric to evaluate regression models, but a high R² can be misleading. Using three model types — mean, linear, and polynomial — the post explains how R² measures only the proportion of variance explained on training data. A mean model yields R²=0 (underfitting), while a high-degree polynomial can achieve R²=1 by memorizing noise (overfitting). The linear model strikes a balance. Because R² doesn't measure predictive performance on unseen data, robustness, or generalization, practitioners should complement it with train-test splits, cross-validation, regularization, adjusted R², and out-of-sample testing.