In statistics, H1 (also written as Ha) stands for the alternative hypothesis. It’s the statement a researcher actually wants to provide evidence for: that a real difference, effect, or relationship exists. H1 is always paired with H0, the null hypothesis, which represents the default position that nothing interesting is happening. Every hypothesis test is essentially a contest between these two.
How H1 and H0 Work Together
Statistical testing starts with an assumption that there’s no effect or no difference. That assumption is H0, the null hypothesis. A typical H0 might say “there is no true mean difference between these two groups.” H1 takes the opposite stance: “there is a true mean difference.”
The key rule is that H0 always contains an equals sign (=, ≥, or ≤), while H1 never does. H1 uses symbols like ≠, >, or <. For example, if you're testing whether the average GPA of college students differs from 2.0, the hypotheses look like this:
- H0: μ = 2.0 (the population mean equals 2.0)
- H1: μ ≠ 2.0 (the population mean does not equal 2.0)
You don’t prove H1 directly. Instead, you collect data and look for enough evidence to reject H0. When you reject the null hypothesis, you adopt the alternative hypothesis. If the evidence isn’t strong enough, you fail to reject H0, but that doesn’t prove H0 is true either.
One-Tailed vs. Two-Tailed H1
The way you write H1 determines whether your test looks for a difference in one specific direction or in both directions.
A two-tailed H1 uses the ≠ symbol. It says “there’s a difference, and I don’t care which direction.” If you’re testing whether a drug changes cholesterol levels compared to a target of 25%, you’d write H1: p ≠ 0.25. The test splits your margin of error evenly between both tails of the distribution. At a standard significance level of 0.05, that means 0.025 sits in each tail.
A one-tailed H1 uses either > or <. It commits to a specific direction. If you want to test whether college students graduate in fewer than five years on average, you'd write H0: μ ≥ 5 and H1: μ < 5. Because all of your error margin is concentrated in one direction, a one-tailed test makes it slightly easier to find significance in that direction.
That easier threshold comes with a rule: you must choose your direction before looking at the data. Switching from a two-tailed test to a one-tailed test after seeing results that almost reached significance is considered inappropriate. Most statistical software defaults to two-tailed p-values for this reason.
Practical Examples of H1
The structure of H1 depends on what question you’re asking. Here are several real-world examples that show the pattern:
- Testing a claim about averages: “The mean number of years Americans work before retiring is 34.” H0: μ = 34, H1: μ ≠ 34
- Testing an upper bound: “Private universities’ mean tuition is more than $20,000 per year.” H0: μ ≤ 20,000, H1: μ > 20,000
- Testing a proportion: “Fewer than 5% of adults ride the bus to work in Los Angeles.” H0: p = 0.05, H1: p < 0.05
- Testing a lower bound: “The chance of developing breast cancer is under 11% for women.” H0: p ≥ 0.11, H1: p < 0.11
Notice how the claim you want to support always goes into H1. The structure forces you to gather evidence against the status quo rather than simply confirming what you already believe.
When H1 Gets Accepted or Rejected
The bridge between your data and a decision about H1 is the p-value. A p-value tells you how likely your observed results would be if H0 were actually true. The smaller the p-value, the harder it is to believe the null hypothesis.
Before running the test, researchers set an alpha level, which is the threshold for how much risk of being wrong they’ll tolerate. The most common alpha is 0.05, meaning a 5% chance of incorrectly rejecting H0. If the p-value falls below alpha, the result is statistically significant and you reject H0 in favor of H1. If the p-value is equal to or greater than alpha, you don’t have enough evidence to support H1.
Errors Connected to H1
Two types of mistakes can happen in this process, and both relate directly to H1.
A Type I error occurs when you reject H0 and accept H1, but H0 was actually true. You’ve concluded there’s an effect when there isn’t one. The probability of this equals your alpha level, so at alpha = 0.05, you accept a 5% chance of this happening.
A Type II error goes the other direction. You fail to reject H0 when H1 is actually true. You’ve missed a real effect. The probability of a Type II error is called beta (β). Statistical power, which is 1 minus beta, measures your test’s ability to correctly detect a real effect and accept H1 when it’s true. A well-designed study typically aims for power of 0.80 or higher, meaning at least an 80% chance of catching a real effect.
H1 in Clinical Trials
In medical research, H1 takes on concrete, practical meaning. A clinical trial testing whether a new treatment outperforms an existing one (called a superiority trial) frames H1 as: “there is a significant difference between the new intervention and the control.” These trials typically use a two-tailed alternative hypothesis, meaning they test whether the new treatment could be either better or worse.
The logic works the same way it does in a classroom example, just with higher stakes. The null hypothesis assumes the new treatment has no advantage. Researchers collect patient data, calculate a test statistic, and check whether the p-value falls below their pre-set significance level. If it does, they reject H0 and conclude the treatment has a real effect. The entire framework exists to guard against mistaking random variation for a genuine medical breakthrough.