Relationship Between F-Test and t-Tests
What you'll learn: How the global F-test and individual t-tests work together to give you a complete picture of your multiple regression model.
Two Tests, Two Questions
In multiple regression, you've learned about two types of hypothesis tests:
- Global F-test: Asks "Does at least one predictor help explain the outcome?"
- Individual t-tests: Ask "Does this specific predictor add value?"
These aren't competing tests—they're complementary tools that answer different questions.
How They Relate
Think of the F-test as a "gatekeeper" test. If your F-test is not significant, it means none of your predictors collectively explain the outcome better than just using the mean. In this case, individual t-tests become less meaningful because the entire model has failed.
But when the F-test is significant, you know at least one predictor matters. Now the t-tests tell you which ones. You might find:
- All predictors significant (rare with many variables)
- Only some predictors significant (common—trim the rest)
- Many predictors insignificant despite significant F-test (suggests multicollinearity or redundant variables)
A Special Case
In simple linear regression (one predictor), the F-test and t-test are mathematically equivalent. In fact, F = t². They're asking the same question: "Does this one predictor matter?" But once you add multiple predictors, they diverge in meaning.
When They Disagree
A significant F-test with all insignificant t-tests often signals multicollinearity—your predictors work together but can't be separated individually.
Key Takeaway: The F-test tells you if your model is worth anything overall; t-tests tell you which predictors deserve to stay. Use both to make informed decisions about your model.