Why Add Multiple Predictors?
What you'll learn: How using multiple predictors instead of just one improves your model's ability to explain real-world variation.
The Single-Variable Limitation
When you build a simple linear regression with one predictor, you're essentially saying "this one factor explains everything about my outcome." But reality is rarely that simple.
For example, predicting house prices using only square footage might give you an R-squared of 0.50—meaning you're explaining only 50% of the variation in prices. What about the other 50%? Location matters. Age of the home matters. Number of bedrooms matters.
Multiple Predictors Capture More Reality
Adding multiple predictors allows your model to account for several factors simultaneously. Each predictor contributes its unique explanatory power to the outcome. When you include square footage and location and age, you're painting a more complete picture.
Think of it like describing why someone is tired: "They slept 4 hours" explains part of it, but adding "and ran a marathon" and "while recovering from flu" explains much more.
Improved Explanatory Power
Multiple regression increases R-squared (the proportion of variance explained) because:
- Each predictor captures different aspects of variation
- You reduce the "unexplained" residual variation
- The model becomes more realistic and useful for prediction
However, adding predictors isn't always better—you need to ensure they contribute meaningfully and don't introduce multicollinearity (which you've already learned about). The goal is parsimony: the simplest model that adequately explains the phenomenon.
Key Takeaway: Adding multiple predictors improves model explanatory power by capturing different sources of variation in the outcome, making your model more aligned with the complexity of real-world relationships.