Course contentsShow
← Data Science
Lesson 621 of 2,14516. Multiple Linear RegressionFree lesson

Interpreting t-Statistics and Confidence Intervals

Understand the t-statistic for each coefficient and construct confidence intervals for population parameters.

Interpreting t-Statistics and Confidence Intervals

You'll learn how to read t-statistics and confidence intervals to judge whether each predictor truly matters in your model.

The Core Idea

After running multiple linear regression, each coefficient gets its own t-statistic and confidence interval. These tell you whether that predictor has a real relationship with the outcome or if its apparent effect could just be random noise.

The t-statistic measures how many standard errors your coefficient estimate is away from zero. Think of it like a signal-to-noise ratio: a large absolute t-value (typically above 2 or below -2) suggests a meaningful signal. The accompanying p-value tells you the probability of seeing such a t-statistic if the true coefficient were actually zero.

Confidence Intervals: A Range of Plausible Values

A 95% confidence interval gives you a range where you're 95% confident the true population coefficient lies. For example, if β₁ = 3.2 with a CI of [1.5, 4.9], you're confident the true effect is somewhere in that range—and importantly, since zero isn't included, the predictor likely matters.

If the confidence interval includes zero, you cannot rule out that the predictor has no effect. This aligns with a non-significant t-test (p-value > 0.05).

Reading the Output Together

When you see a coefficient with t = 4.5 and CI = [2.1, 6.3], both tell the same story: this predictor is statistically significant. The t-statistic gives you the test result; the confidence interval shows you the precision and magnitude of the estimate.

Key Takeaway: The t-statistic tests if a coefficient differs from zero; the confidence interval shows the range of plausible values—both help you decide if a predictor truly contributes to your model.