Choosing Between z-Tests and Other Tests
You'll learn when z-tests are the right tool and when you need to reach for alternatives.
The Core Idea
Not every hypothesis test should use a z-test. The z-test is powerful but comes with specific requirements—think of it like a precision instrument that only works under the right conditions. Choosing the wrong test is like using a wrench when you need a screwdriver: you might get somewhere, but the results won't be reliable.
When z-Tests Are Appropriate
A z-test is your go-to when all three conditions align:
- Known population standard deviation (σ): You need to know σ, not just estimate it from your sample
- Large sample size: Typically n ≥ 30, so the Central Limit Theorem ensures your sampling distribution is approximately normal
- Normal population OR large n: Either the population itself is normally distributed, or your sample is large enough that the CLT kicks in
When to Choose Other Tests
If any z-test condition fails, you need an alternative:
- σ unknown and small sample: Use a t-test instead (coming in your future lessons). The t-distribution accounts for the extra uncertainty from estimating σ
- Very small samples from non-normal populations: Non-parametric tests may be needed
- Testing variances rather than means: You'll need chi-square or F-tests
Think of it as a decision tree: start by asking "Do I know σ?" and "Is my sample large enough?" Your answers determine the path forward.
Key Takeaway: Use z-tests when you know the population standard deviation and have either a large sample (n ≥ 30) or a normally distributed population; otherwise, different tests account for the additional uncertainty or distributional concerns.