P-Value Calculation Differences
What You'll Learn
You'll learn how the same test statistic produces different p-values depending on whether you're running a one-tailed or two-tailed test.
The Core Idea
Imagine you calculate a test statistic (like a z-score or t-statistic) and get a value of 2.1. The p-value tells you how extreme this result is under the null hypothesis. But "extreme" means different things depending on your alternative hypothesis:
-
Two-tailed test: You care about differences in either direction (bigger or smaller). You look at both tails of the distribution. If your test statistic is 2.1, you find the probability of getting ≥2.1 or ≤-2.1. This doubles the area in the tail.
-
One-tailed test: You only care about one specific direction (say, bigger). You look at just one tail. For the same test statistic of 2.1, you only find the probability of getting ≥2.1. This is exactly half the two-tailed p-value.
The Relationship
For the same test statistic:
p-value (two-tailed) = 2 × p-value (one-tailed)
Why This Matters
If your test statistic gives a one-tailed p-value of 0.025, the two-tailed p-value is 0.05. This affects your decision:
- At α = 0.05, the two-tailed test would reject (p = 0.05)
- At α = 0.025, the one-tailed test would be borderline
You're dividing your significance level across both tails in a two-tailed test, making it slightly harder to reject the null compared to a one-tailed test in the predicted direction.
Key Takeaway: The same test statistic yields different p-values based on test directionality—two-tailed p-values are always twice the corresponding one-tailed p-value because they account for extremity in both directions.