Course contentsShow
← Data Science
Lesson 608 of 2,14516. Multiple Linear RegressionFree lesson

Standardized Coefficients (Beta Weights)

How to calculate and interpret standardized coefficients to compare relative importance across predictors.

Standardized Coefficients (Beta Weights)

What You'll Learn

You'll learn how to calculate and interpret standardized coefficients to fairly compare which predictors have the strongest effect on your outcome.

The Problem with Raw Coefficients

In multiple linear regression, raw coefficients tell you how much your outcome changes per one-unit increase in a predictor. But here's the catch: what if one predictor is measured in years (ranging 20–60) while another is measured in thousands of dollars (ranging 20,000–80,000)? A coefficient of 2.0 for age and 0.5 for income doesn't mean age is "more important"—the units are completely different.

What Are Standardized Coefficients?

Standardized coefficients (also called beta weights or β weights) put all predictors on the same scale by expressing them in standard deviation units. This lets you compare apples to apples.

When you standardize:

  • Each predictor is converted to z-scores (mean = 0, SD = 1)
  • The outcome is also converted to z-scores
  • The resulting coefficients tell you: "If this predictor increases by 1 standard deviation, the outcome changes by β standard deviations"

How to Calculate

You can either:

  1. Manually standardize all variables before fitting regression
  2. Calculate from raw coefficients: β_standardized = β_raw × (SD_x / SD_y)

Interpretation

If age has β = 0.6 and income has β = 0.3, age has twice the effect size of income. A 1-SD increase in age produces a 0.6-SD change in the outcome, holding other variables constant.

Key Takeaway: Standardized coefficients remove the effect of measurement units, allowing you to directly compare the relative importance of predictors—the larger the absolute value of β, the stronger the predictor's influence.