So what is the null hypothesis? It's the claim that no effect or difference exists, and you treat it as true until your data proves otherwise. Its counterpart, the alternative (Hₐ), argues the reverse, and it's the statement a researcher usually wants to back up. None of that is new to you, probably. The tricky part surfaces somewhere else, namely, when plain research questions have to become symbols and exact wording. Operator, tail, overall form - all three have to be right. Miss one of them and the whole test breaks on the spot.
For this article, I pulled together a set of null and alternative hypothesis examples and set each verbal claim beside its maths. The samples come from eight fields, and you'll recognize every one of them.
Quick Reference: H₀ vs. Hₐ Formula Cheat Sheet
.webp)
Do enough hypothesis testing, and you start to see the same underlying structure that keeps coming back. The table below lines up all eight scenarios so that the structure is easy to spot. One rule holds with no exceptions. Equality always is in the null (=, ≤, ≥). The alternative takes the strict form only (≠, <, >), and an equals sign is never allowed there. Keep this table close when you work through null vs alternative hypothesis examples:
Rows 3 and 6 read almost the same and both point rightward because the researcher already expects an increase, so the direction is already implied by the alternative. Row 8 is the exception. ANOVA runs on an F-distribution, which makes the test right-tailed on its own, even though the hypothesis wording stays non-directional and only asks if any single group differs from the others.
Deadlines Don't Scare Us
Get a polished, custom homework written by pros who know your subject inside and out.
Null Hypothesis (H₀) Examples
There's one check a null hypothesis has to clear before it counts as valid, and it's non-negotiable: equality has to be somewhere within it. Only three symbols qualify here: =, ≤, and ≥. When not one of them shows up, the statement is broken.
Why does this matter so much? Look at what H₀ is really for. It pins down the status quo, that "nothing is happening here" claim a study exists to knock down, which means it needs a fixed value to aim at. Each null hypothesis example below runs on its own research setup, and every one keeps the wording and the notation together so you can watch one turn into the other.
Null Hypothesis Example 1: Single Population Mean
Field: Manufacturing Quality Control
Picture a plant that fills cereal boxes to a labeled weight of 500 grams. Quality control wants evidence that the line hasn't drifted. So the null takes the optimistic position: the true mean weight across all boxes is exactly 500 grams.
- Wording: The average fill weight of the production line equals 500 grams.
- Notation: H₀: μ = 500
That equals sign carries the whole test. A box that runs heavy counts against H₀ just as much as one that runs light, and that two-sided sensitivity is what makes the setup two-tailed.
Null Hypothesis Example 2: Comparing Two Means
Field: Education & Teaching Methods
A school pits a new project-based method against its usual lecture format. Two groups of students, one shared exam at the end, one comparison to run. What the null claims is that the method changed nothing about average performance.
- Wording: The mean exam score of the project-based group equals the mean score of the lecture group.
- Notation: H₀: μ₁ = μ₂
There's a second way to write it that some instructors prefer, which pins the difference between the two means at zero: H₀: μ₁ − μ₂ = 0. Same claim, different clothing.
Null Hypothesis Example 3: Single Proportion
Field: E-Commerce Conversion Rates
An online store rebuilds its checkout page and hopes a larger share of visitors finish a purchase. The old page converted 4%. The team already predicts an improvement, so the null holds the line at the old rate or anything under it.
- Wording: The conversion rate of the new checkout page is at most 4%.
- Notation: H₀: p ≤ 0.04
That ≤ still satisfies the equality rule, and it sets up a one-tailed test pointing right.
Null Hypothesis Example 4: Comparing Two Proportions
Field: Clinical Drug Trials
A trial hands one group an experimental drug and a second group a placebo, then counts recoveries in each. The question on the table is a difference in recovery proportions. The null denies that any such difference exists.
- Wording: The recovery rate in the drug group equals the recovery rate in the placebo group.
- Notation: H₀: p₁ = p₂
If you line up several examples of null hypothesis statements side by side, this shape keeps coming back: a flat claim of no difference, anchored by an equals sign.
Null Hypothesis Example 5: Correlation Between Two Variables
Field: Environmental Science
An environmental team samples a lake to find out if nitrogen runoff tracks with algae growth. The parameter that matters is the population correlation coefficient, ρ. What the null asserts is that no linear relationship connects the two.
- Wording: There is no linear relationship between nitrogen runoff and algae growth.
- Notation: H₀: ρ = 0
A ρ of exactly zero means the two measurements don't rise and fall together in any straight-line way. Reject that, and a real association is on the table.
Null Hypothesis Example 6: One-Tailed Directional Test
Field: Employee Training Programs
A company runs a workshop and bets it will lift the number of tickets each support agent closes in a day. Output before the session averaged 50 tickets. When the prediction has a direction, the null has to cover the opposite direction plus the baseline.
- Wording: The training does not push average output above 50 tickets a day; the mean stays at 50 or below.
- Notation: H₀: μ ≤ 50
The pick between ≤ and ≥ hangs entirely on which way the claim leans. Here, the alternative points up, so the null swallows everything at or beneath the starting figure.
Null Hypothesis Example 7: Two-Tailed Non-Directional Test
Field: Psychology & Behavioral Study
A researcher clocks reaction time after participants look at a mildly stressful image, then holds the result up against a known baseline of 250 milliseconds. No guess about direction here, only a question of change. The null keeps the mean nailed to the baseline.
- Wording: The stimulus has no effect on mean reaction time, which holds at 250 milliseconds.
- Notation: H₀: μ = 250
Faster or slower, any shift argues against this null. That openness on both sides is exactly what marks a non-directional test, and it's why the equals sign shows up instead of an inequality.
Null Hypothesis Example 8: Multi-Group Comparison / ANOVA
Field: Agricultural Crop Yields
An agronomist tries three fertilizers on identical plots and records each plot's yield. One test now has to weigh all three groups at the same time, and that job belongs to ANOVA. The null says that every fertilizer lands on the same average yield.
- Wording: The three fertilizers produce equal mean crop yields.
- Notation: H₀: μ₁ = μ₂ = μ₃
A single chain of equals signs handles all the groups at once. By this stage, most people reach for statistics software such as R, SPSS, or Excel, because the arithmetic turns brutal by hand the moment you clear two groups. Reject this null, and you learn that at least one fertilizer stands apart, though the test won't point a finger at which.
Alternative Hypothesis (Hₐ) Examples
The alternative hypothesis is the claim you're actually chasing. It's the reason the study exists, the statement that says something is going on: an effect, a gap, a link. And it has a strict dress code, the exact opposite of the null's. Equality is banned. No equals sign, no ≤, no ≥. What Hₐ uses instead are the sharp operators, ≠, <, and >. Every alternative hypothesis example below flips a null from the previous section into the claim a researcher hopes the data will back, and each one shows the wording next to the notation.
Alternative Hypothesis Example 1: Single Population Mean
Field: Manufacturing Quality Control
Back to the cereal line and its 500-gram target. Quality control doesn't just want reassurance, it wants to catch a problem if one exists. The alternative says the line has slipped off target in some direction, either heavy or light.
- Wording: The average fill weight of the production line is not equal to 500 grams.
- Notation: Hₐ: μ ≠ 500
That ≠ leaves both doors open, above and below, which is the signature of a two-tailed test. No commitment to which way the drift runs, only that it's there.
Alternative Hypothesis Example 2: Comparing Two Means
Field: Education & Teaching Methods
The school still wants to know if its project-based method beats plain lectures. Here the researchers aren't predicting a winner ahead of time, only that the two methods differ. So the alternative claims the average scores aren't the same.
- Wording: The mean exam score of the project-based group differs from the mean score of the lecture group.
- Notation: Hₐ: μ₁ ≠ μ₂
You could also write it as Hₐ: μ₁ − μ₂ ≠ 0. This is a non-directional example of alternative hypothesis phrasing, since it allows either group to come out on top.
Alternative Hypothesis Example 3: Single Proportion (E-Commerce Conversion Rates)
The store had a specific hope for its new checkout page: more purchases, not fewer. That expectation was baked in from the start, so the alternative points one way only.
- Wording: The conversion rate of the new checkout page is greater than 4%.
- Notation: Hₐ: p > 0.04
The > makes this directional. The team isn't asking if the rate merely changed; they're asking if it climbed. A drop, or no movement at all, would leave the null standing.
Alternative Hypothesis Example 4: Comparing Two Proportions
Field: Clinical Drug Trials
The drug trial compared recoveries between a treatment group and a placebo group. Without a prior claim about which should do better, the alternative just asserts the two rates aren't equal.
- Wording: The recovery rate in the drug group differs from the recovery rate in the placebo group.
- Notation: Hₐ: p₁ ≠ p₂
Two-tailed again. In real trials you'll sometimes see a directional version instead (Hₐ: p₁ > p₂) when the goal is proving the drug is strictly better, so the wording of the research question decides the operator.
Alternative Hypothesis Example 5: Correlation Between Two Variables
Field: Environmental Science
The environmental team suspected nitrogen runoff and algae growth were connected somehow. The alternative doesn't specify positive or negative, only that a real linear relationship is there.
- Wording: There is a linear relationship between nitrogen runoff and algae growth.
- Notation: Hₐ: ρ ≠ 0
Any ρ that isn't zero supports this, whether the two variables rise together or one climbs as the other falls. That's what keeps it non-directional.
Alternative Hypothesis Example 6: One-Tailed Directional Test
Field: Employee Training Programs
The workshop was supposed to raise the tickets each agent closes per day, with 50 as the pre-training average. The whole point was an increase, so the alternative commits to a direction without apology.
- Wording: The training raises average output above 50 tickets a day.
- Notation: Hₐ: μ > 50
This is the clearest directional alternative hypothesis example in the set. The > matches the ≤ from its paired null, and together they cover every possible value with no overlap.
Alternative Hypothesis Example 7: Two-Tailed Non-Directional Test
Field: Psychology & Behavioral Study
The reaction-time study measured whether a stressful image changed response speed against a 250-millisecond baseline. No prediction about faster or slower was made, just a question of any change. The alternative reflects that openness.
- Wording: The stimulus affects mean reaction time, which no longer equals 250 milliseconds.
- Notation: Hₐ: μ ≠ 250
Speeding up or slowing down both count as support. The ≠ is doing the same both-directions job you saw in Example 1, which is why psychologists lean on this form when they have no strong theory about direction yet.
Alternative Hypothesis Example 8: Multi-Group Comparison / ANOVA
Field: Agricultural Crop Yields
Three fertilizers, identical plots, one ANOVA weighing them all together. The null said every mean yield matched. The alternative can't just flip that to a neat inequality, because "not all equal" covers a lot of possibilities.
- Wording: At least one fertilizer produces a different mean yield from the others.
- Notation: Hₐ: not all μᵢ are equal
Writing Hₐ: μ₁ ≠ μ₂ ≠ μ₃ is a mistake, since the real claim only needs one group to stand apart, not all three to differ from each other. A significant result tells you a difference exists somewhere, and a follow-up post hoc test is what actually locates it.
Weak vs. Strong Null and Alternative Hypothesis Examples
You can run a test perfectly and get nothing usable, and the fault usually starts at the hypothesis: an undefined variable, an outcome you can't measure, a null and alternative that don't fit. The software won't complain. Below are common examples of null and alternative hypothesis writing that break, and their fixes.
Weak Example: "Students who study more do better in school."
- What's wrong: "Study more" has no unit, "do better" - no yardstick, and "school" names no specific group.
- Improved H₀: Mean GPA is equal for students who study 10+ hours a week and those who study fewer. (H₀: μ₁ = μ₂)
- Improved Hₐ: Mean GPA differs between the two groups. (Hₐ: μ₁ ≠ μ₂)
Weak Example: "The new app makes people happier."
- What's wrong: "Happier" isn't measurable as phrased, and no population is defined. A test needs an operational stand-in, like a score on a validated scale.
- Improved H₀: Mean well-being score of app users equals 60 on the standardized index. (H₀: μ = 60)
- Improved Hₐ: Mean well-being score of app users is greater than 60. (Hₐ: μ > 60)
Weak Example: "H₀: the diet reduces weight. Hₐ: the diet increases weight."
- What's wrong: These two don't fit. A proper null and alternative must split every outcome with no gap or overlap, and this pair skips "no change" while both claim a direction.
- Improved H₀: Mean weight loss on the diet is 0 kg or less. (H₀: μ ≤ 0)
- Improved Hₐ: Mean weight loss on the diet is greater than 0 kg. (Hₐ: μ > 0)
Weak Example: "The training program affects everything about employee performance."
- What's wrong: The scope is too broad. Pick one variable and define it.
- Improved H₀: Mean weekly sales equal $8,000 for trained employees. (H₀: μ = 8000)
- Improved Hₐ: Mean weekly sales do not equal $8,000 for trained employees. (Hₐ: μ ≠ 8000)
Weak Example: "H₀: the fertilizer has no effect on tomato plants. Hₐ: the fertilizer changes soil pH."
- What's wrong: The null is about the plant effect, the alternative about the soil pH, two different studies. Both must describe the same variable in the same population.
- Improved H₀: Mean tomato yield with the fertilizer equals the mean yield without it. (H₀: μ₁ = μ₂)
- Improved Hₐ: Mean tomato yield with the fertilizer differs from mean yield without it. (Hₐ: μ₁ ≠ μ₂)
Weak Example: "The medication changes recovery time, maybe faster, maybe slower, we'll see."
- What's wrong: A researcher must settle it before collecting data, since that fixes the operator and tail in advance.
- Improved H₀: Mean recovery time on the medication equals 7 days. (H₀: μ = 7)
- Improved Hₐ: Mean recovery time on the medication does not equal 7 days. (Hₐ: μ ≠ 7)
If all of this proves too difficult for you to handle in a single all-nighter, you can always ask us, ‘Do my statistics homework for me, please.’
Skip the Stats Struggle
Our experts handle your hardest assignments while you focus on taking a little breather.
Final Thoughts
Everything here comes down to two opposing statements. The null keeps equality (=, ≤, ≥) and claims nothing is happening. The alternative drops equality and makes the research claim (≠, <, >). Get those right, then check every hypothesis for a measurable variable, a defined population, matched statements, and clear direction.
