What Is a Null Hypothesis and How Do You Write One

What Is a Null Hypothesis? The Definition Every Student Needs

A null hypothesis is a formal statistical statement asserting that there is no significant effect, no meaningful difference, and no real relationship between the variables being studied. It is written as H0 (pronounced “H-naught” or “H-zero”) and serves as the default starting point for any hypothesis test. When you run an experiment or statistical analysis, you are always testing whether your data provides enough evidence to reject this default assumption.

Think of it this way. A criminal trial begins with the presumption of innocence — “not guilty” is the default until the evidence proves otherwise. The null hypothesis works exactly the same way. You start from a position of “nothing is happening here” and demand that your data demonstrate otherwise before you change that position. This is why the null hypothesis is sometimes described as the “status quo” claim. Understanding the full hypothesis testing framework is essential before you can write a null hypothesis correctly.

The word “null” comes from the Latin nullus, meaning “none” or “nothing.” In statistics, it signifies that the presumed effect is zero. If you are testing whether a drug lowers blood pressure, the null hypothesis states that the drug has zero effect on blood pressure. If you are testing whether two groups have different mean scores, the null hypothesis states the difference in means is zero. Understanding p-values and significance levels is how you determine whether your data provides enough evidence to reject this zero-effect assumption.

Why the Null Hypothesis Exists: The Logic of Falsifiability

The null hypothesis exists because of a fundamental principle of scientific reasoning: you cannot prove something is true, but you can disprove it. This principle — called falsifiability — was formalized by philosopher Karl Popper and is the cornerstone of the scientific method. Exploring the scientific method in depth helps clarify why falsifiability is so central to research design.

By setting up a null hypothesis that assumes no effect, researchers create a falsifiable claim. If the data consistently contradicts the null — if blood pressure consistently drops, if test scores consistently differ — that contradiction accumulates into statistical evidence strong enough to reject the null. You haven’t proven the alternative true. You’ve only shown the null is implausible given your data. This careful logic is what keeps scientific conclusions honest and prevents researchers from overclaiming. Understanding statistical misuse and p-hacking shows what happens when this logic is violated.

Who Formalized the Null Hypothesis?

The null hypothesis as a formal statistical concept was developed through the work of three towering figures in 20th-century statistics. Ronald A. Fisher — a British statistician and geneticist who worked at institutions including Rothamsted Experimental Station in England and later University of Adelaide in Australia — introduced the concept of testing against a null hypothesis in his landmark 1925 work Statistical Methods for Research Workers. Fisher insisted the null must be exact and unambiguous because it forms the mathematical basis for calculating probability.

Jerzy Neyman and Egon Pearson later extended and formalized Fisher’s framework. Their 1933 paper in Philosophical Transactions of the Royal Society A introduced the concept of the alternative hypothesis and the formal decision-making framework of hypothesis testing that researchers use today. The Neyman-Pearson approach frames the null as one of two competing hypotheses — H0 and H1 — and defines Type I and Type II errors as the two possible mistakes. Understanding Type I and Type II errors in full context is essential for anyone working with hypothesis tests in research or data science.

Null Hypothesis vs. Alternative Hypothesis: What’s the Difference?

Every hypothesis test involves two competing statements: the null hypothesis (H0) and the alternative hypothesis (H1 or Ha). Understanding their relationship is essential because they are not just opposites — they serve completely different functions in research design and statistical reasoning. The null hypothesis is what you test against. The alternative hypothesis is what you’re trying to demonstrate.

The null and alternative hypotheses are exhaustive (together, they cover every possible outcome) and mutually exclusive (only one can be true at a time). This is not a coincidence — it is a deliberate design feature of the hypothesis testing framework. It ensures that every possible outcome of your study is accounted for. The difference between qualitative and quantitative research is also relevant here, since hypothesis testing in the traditional sense applies specifically to quantitative, measurable variables.

Why You Test the Null, Not the Alternative

This is one of the most counterintuitive things about hypothesis testing for students who encounter it for the first time. You’re interested in the alternative hypothesis — you want to show that your treatment works, that the two groups differ, that the relationship is real. So why do you test the null instead?

The answer is probability. It is much easier to calculate the probability of observing data given a specific, defined baseline (the null) than to calculate probabilities against a vague claim of “some effect exists.” The null hypothesis gives you a precise mathematical starting point. The p-value — the cornerstone of hypothesis testing — is the probability of getting results as extreme as yours if the null were true. You can compute this precisely. You cannot compute the probability of results “given some unspecified effect exists.” Understanding probability distributions is the mathematical foundation for why this works.

One Research Question, Two Hypotheses: A Worked Example

Suppose a psychology researcher at Harvard University wants to study whether mindfulness meditation reduces anxiety in college students. The research question is: “Does an 8-week mindfulness program reduce anxiety scores in college students compared to a control group?”

From this single research question, two hypotheses immediately follow. The null hypothesis states: H0: There is no difference in mean anxiety scores between students who completed the mindfulness program and those who did not (μ_treatment = μ_control). The alternative hypothesis states: Ha: Students who completed the mindfulness program have lower mean anxiety scores than those who did not (μ_treatment < μ_control). Notice that the alternative here is directional — the researcher specifically predicts lower anxiety in the treatment group, making this a one-tailed test. If the researcher simply expected a difference (in either direction), they would write Ha: μ_treatment ≠ μ_control (two-tailed). Understanding the one-sample t-test builds directly on this framework when you have a single group mean to compare against a known value.

How to Write a Null Hypothesis: A Step-by-Step Process

Writing a correct null hypothesis is a skill that takes practice — but it follows a clear, repeatable process. Once you know the six steps below, you can write a null hypothesis for any research question, in any discipline, using any statistical test. Mastering academic writing for research papers starts here, at the level of hypothesis formulation.

Null Hypothesis Examples Across Disciplines

The null hypothesis appears in every field that uses data — psychology, medicine, education, business, biology, and more. The structure is always the same: no effect, no difference, no relationship. But the specific form it takes depends on your research context, the type of data you’re working with, and the statistical test you’ll use. Understanding descriptive vs. inferential statistics is fundamental here — null hypothesis testing is always inferential, because you’re drawing conclusions about a population from sample data.

Psychology and Behavioral Science Examples

In psychology research conducted at institutions like the American Psychological Association (APA)-affiliated universities, null hypotheses typically compare group means or test correlations between psychological constructs.

Example 1 — Comparing treatment groups: A researcher wants to know whether cognitive-behavioral therapy (CBT) reduces depression scores more effectively than standard counseling in college students. The null hypothesis is: H0: There is no difference in mean depression scores between the CBT group and the standard counseling group (μ_CBT = μ_standard). The alternative: Ha: The CBT group shows lower mean depression scores than the standard counseling group (μ_CBT < μ_standard).

Example 2 — Testing a correlation: A researcher investigates whether social media usage is related to self-esteem scores among university students. Null hypothesis: H0: There is no correlation between daily social media use and self-esteem scores in the population (ρ = 0). Alternative: Ha: There is a negative correlation between daily social media use and self-esteem scores (ρ < 0). Mastering correlation and statistical relationships is the natural next step after you’ve formulated this kind of null hypothesis.

Medical and Clinical Research Examples

In clinical trials — the gold standard of medical research — the null hypothesis is typically that a new treatment performs no better than a placebo or an existing standard of care. The Food and Drug Administration (FDA) in the United States requires clinical trials to formally test null hypotheses before approving new medications, making null hypothesis testing a legal and regulatory requirement in medicine.

Example 3 — Drug efficacy trial: A pharmaceutical company tests whether a new antihypertensive drug reduces systolic blood pressure more effectively than a placebo. Null hypothesis: H0: The mean reduction in systolic blood pressure is the same in the drug group and the placebo group (μ_drug = μ_placebo). Alternative: Ha: The mean reduction in systolic blood pressure is greater in the drug group than in the placebo group (μ_drug > μ_placebo). The CONSORT standards for clinical trials require explicit null hypothesis statements in all published trial protocols.

Education Research Examples

Education researchers at institutions like the American Educational Research Association (AERA) use null hypotheses to evaluate teaching methods, interventions, and policy changes.

Example 4 — Teaching method comparison: A professor at a large US public university wants to know whether flipped classroom instruction improves exam performance compared to traditional lecture. Null hypothesis: H0: The mean exam scores of students in the flipped classroom format equal those in the traditional lecture format (μ_flipped = μ_traditional). Alternative: Ha: The mean exam scores differ between the two formats (μ_flipped ≠ μ_traditional). This is two-tailed because the researcher is open to either format performing better. Comparing online and in-person learning outcomes is exactly the kind of research question that a well-formulated null hypothesis can help answer rigorously.

Business and Marketing Examples

In business analytics and A/B testing — a standard practice at companies like Google, Amazon, and virtually every data-driven organization — the null hypothesis is tested hundreds of times daily. Every product test, button color experiment, and pricing change is evaluated against a null.

Example 5 — A/B testing a website: A marketing team wants to know whether changing a call-to-action button from blue to green increases click-through rates. Null hypothesis: H0: The click-through rate is the same for the blue button and the green button (p_blue = p_green). Alternative: Ha: The click-through rates differ (p_blue ≠ p_green). This is a test of proportions. Chi-square tests for independence are frequently used for this kind of proportion-based null hypothesis testing in marketing research.

How Hypothesis Testing Works: From Null Hypothesis to Decision

Writing the null hypothesis is only the beginning. The entire hypothesis testing process flows logically from formulating H0 to making a final statistical decision. Understanding each step — and how they connect — is what separates students who get hypothesis testing right from those who make procedural errors that invalidate their results. The comprehensive hypothesis testing guide covers all steps in depth; here we focus specifically on the null hypothesis’s role in each stage.

Step 1: State the Hypotheses

Both H0 and Ha must be written before data collection. This is not optional — it is the rule that prevents “p-hacking,” the problematic practice of adjusting hypotheses after seeing results. Writing your hypotheses first commits you to a specific test and direction, making your conclusions credible. The American Psychological Association’s journal guidelines require pre-registration of hypotheses for this reason. Statistical misuse and p-hacking are serious problems in published research, and pre-specifying your null hypothesis is the primary defense against them.

Step 2: Choose a Significance Level (Alpha)

Alpha (α) is the threshold you set for rejecting the null hypothesis. The most common choice is α = 0.05, meaning you are willing to accept a 5% chance of incorrectly rejecting a true null hypothesis (a Type I error). Some fields use α = 0.01 (medicine, where false positives have serious consequences) or α = 0.10 (exploratory social science research). You must choose alpha before collecting data — choosing it after is a form of manipulation. The relationship between p-values and significance levels is one of the most important concepts in applied statistics.

Step 3: Collect Data and Calculate the Test Statistic

Once your hypotheses are stated and your alpha is set, collect your sample data and calculate the appropriate test statistic. The test statistic depends on your hypothesis type: t-tests for comparing means, z-tests for proportions, F-tests in ANOVA for multiple groups, chi-square tests for categorical data, and correlation tests for relationships between continuous variables. The test statistic measures how far your sample result is from the null hypothesis value, in units of standard error. The complete guide to t-tests walks through how to calculate and interpret this test statistic for mean comparisons.

Step 4: Compute the P-Value

The p-value is the probability of observing a test statistic as extreme as, or more extreme than, the one you calculated — assuming the null hypothesis is true. A small p-value means your data would be very unlikely if the null were true. It is not the probability that the null is true. This distinction matters enormously. A p-value of 0.03 means “if there were really no effect, we would see results this extreme only 3% of the time” — not “there is a 3% chance the null is true.” Understanding p-values correctly protects you from one of the most pervasive misinterpretations in statistical reporting.

Step 5: Compare P-Value to Alpha and Make a Decision

This is the decision point. If p < α, you reject the null hypothesis and conclude that the data provide sufficient evidence to support the alternative hypothesis. If p ≥ α, you fail to reject the null hypothesis. Note the precise language: you “fail to reject” — not “accept” — the null. Failing to reject does not mean the null is true; it means you don’t have enough evidence to rule it out. This is the conclusion from Penn State’s STAT 200 program, one of the most widely referenced introductory statistics curricula in US universities.

Step 6: Report the Conclusion in Context

Statistical conclusions must be translated back into the context of the original research question. Don’t just report “p < 0.05, reject H0.” Say what that means for your study: “The results suggest that students who completed the mindfulness program showed significantly lower anxiety scores than the control group (t(48) = −2.84, p = 0.006), providing evidence that the 8-week program reduces anxiety.” Always report effect size alongside the p-value — statistical significance is not the same as practical significance. A tiny effect can be statistically significant with a large enough sample size. Transparent reporting of statistical results is a core expectation in academic research.

Type I and Type II Errors: When Hypothesis Testing Goes Wrong

No hypothesis test is perfect. Because you are making inferences about a population from a sample, there is always a chance of error. The null hypothesis framework defines exactly two types of errors — and understanding them is crucial for evaluating the quality and reliability of any research. Mastering Type I and Type II errors is essential for any statistics, research methods, or experimental design course.

Type I Error: The False Positive

A Type I error occurs when you reject a null hypothesis that is actually true. You conclude an effect exists when it doesn’t. This is a false positive. In medical research, a Type I error might mean concluding a useless drug works. In education, it might mean rolling out a new teaching method that actually provides no benefit.

The probability of making a Type I error is exactly equal to your significance level alpha. If α = 0.05, you accept a 5% chance of incorrectly rejecting a true null hypothesis across repeated experiments. This is why researchers in high-stakes fields like medicine often use α = 0.01 or even α = 0.001 — to reduce the risk of false positives. The replication crisis in psychology and other social sciences — where many published findings failed to replicate — is partly attributable to widespread use of α = 0.05 without adequate power, leading to inflated Type I error rates across the literature. The role of p-hacking in the replication crisis explores this in detail.

Type II Error: The False Negative

A Type II error occurs when you fail to reject a null hypothesis that is actually false. A real effect exists — but your study missed it. This is a false negative. The probability of making a Type II error is called beta (β). The statistical power of a test is 1 − β: the probability of correctly detecting a real effect when one exists.

Type II errors are commonly caused by small sample sizes (underpowered studies), high measurement noise, or too-stringent alpha thresholds. A study that fails to find an effect is not proof that no effect exists — it may simply mean the study wasn’t sensitive enough to detect it. Statistical power analysis and Cohen’s d are the tools researchers use before data collection to ensure their study has enough power to detect the effects they care about.

Balancing Type I and Type II Errors

Here’s the uncomfortable truth: reducing Type I error (by lowering alpha) automatically increases the risk of Type II error, and vice versa. They are in tension. Setting α = 0.001 to minimize false positives makes it much harder to reject the null even when a real effect exists — increasing β and reducing power. This trade-off is inherent to the hypothesis testing framework and there is no perfect solution. Conducting a proper power analysis before data collection is the standard approach for navigating this trade-off responsibly — it allows you to choose an alpha level and target power that is appropriate for your study’s stakes and context.

One-Tailed vs. Two-Tailed Null Hypotheses: When to Use Each

One of the most common decision points students face when writing a null hypothesis is whether to use a one-tailed or two-tailed test. This decision determines the shape of your alternative hypothesis and affects how you interpret the p-value. It must be made before data collection — never after — based on what your theory and prior research predict. The statistical theory behind sampling distributions explains mathematically why this choice affects your p-value and critical values.

Two-Tailed Tests (Non-Directional)

A two-tailed null hypothesis predicts no difference in either direction. The alternative hypothesis is non-directional: “the groups differ, but I don’t specify which direction.” Use a two-tailed test when you have no strong theoretical or empirical reason to predict the direction of the effect — or when the consequences of effects in either direction are equally important to detect.

Two-tailed tests are the conservative default in most academic disciplines. H0: μ1 = μ2; Ha: μ1 ≠ μ2. The critical region is split across both tails of the distribution, and the p-value represents the probability of results as extreme as yours in either direction. Most published psychology, education, and social science research uses two-tailed tests unless a specific directional prediction is theoretically justified. The z-score table shows how critical values differ for one- and two-tailed tests at the same alpha level.

One-Tailed Tests (Directional)

A one-tailed null hypothesis predicts that any difference will be in a specific direction. Use a one-tailed test only when you have a strong prior theoretical basis for predicting the direction of the effect, and when detecting an effect in the opposite direction would be scientifically meaningless.

For example, if you’re testing whether a new study technique improves test scores — not decreases them — you could write: H0: μ_new ≤ μ_standard; Ha: μ_new > μ_standard. One-tailed tests have more statistical power to detect effects in the predicted direction (the critical region is concentrated in one tail), but they are criticized for being inappropriate when used to justify choosing whichever tail gives the desired result after seeing data. The t-distribution table shows the different critical values for one- and two-tailed tests, which is a practical reference for any hypothesis-testing assignment.

Common Mistakes Students Make When Writing the Null Hypothesis

Hypothesis testing errors are surprisingly consistent across courses and institutions. The same mistakes appear on assignments at MIT, UCLA, University of Edinburgh, and community colleges alike. Knowing these traps in advance is the most efficient way to avoid them. Understanding common academic writing mistakes applies just as much to statistical writing as to essays.

Mistake 1 — Writing the Null as the Research Hypothesis

The most common error: students write the hypothesis they are actually trying to support as the null hypothesis. The null is always the “no effect” claim — it is what you’re testing against, not what you’re hoping to show. If you believe that exercise improves mental health, your null is: “Exercise has no effect on mental health outcomes.” Your alternative is: “Exercise improves mental health outcomes.” Getting these swapped around invalidates your entire test.

Mistake 2 — Writing the Null Without an Equality Sign

A null hypothesis like “H0: μ > 50” or “H0: μ1 ≠ μ2” is mathematically invalid. The null must contain =, ≤, or ≥. Without equality, there is no specific baseline value to build the test distribution around. No equality sign means no p-value can be properly calculated. Attention to technical precision in academic writing is just as important in statistics notation as in prose grammar.

Mistake 3 — Using Sample Statistics Instead of Population Parameters

Hypotheses are always about populations, never about samples. Writing “H0: x̄ = 50” (using x̄, the sample mean) instead of “H0: μ = 50” (using μ, the population mean) is a fundamental conceptual error. Your sample is just the evidence you use to make inferences — the hypothesis is about the underlying population. This distinction between sample statistics and population parameters is central to inferential statistics as a discipline.

Mistake 4 — Saying “We Accept the Null Hypothesis”

When your results are not statistically significant, you “fail to reject” the null hypothesis — you never “accept” it. Accepting the null implies you have proven there is no effect, which you cannot do from a single study. Failing to reject merely means the evidence was insufficient to rule out the null. This matters enormously in scientific communication: “we failed to find evidence of an effect” is very different from “we proved there is no effect.” Writing precise, accurate scientific sentences is a skill that prevents this kind of language error.

Mistake 5 — Choosing a One-Tailed Test After Seeing the Data

Switching from a two-tailed to a one-tailed test after observing which direction the data went is a form of data manipulation — even if unintentional. It inflates the Type I error rate and produces unreliable results. Your choice of test direction must be justified by theory before data collection and documented in your methods section. Research pre-registration — now standard practice at journals like Psychological Science and on platforms like the Open Science Framework (OSF) — exists specifically to prevent this. The full impact of p-hacking on research reliability is explored in our dedicated guide.

Mistake 6 — Confusing Statistical Significance With Practical Importance

A statistically significant result (p < α) does not automatically mean the effect is large, important, or meaningful in practice. With a large enough sample, even a trivially small difference becomes statistically significant. Always report an effect size measure alongside your p-value — Cohen’s d for mean differences, r for correlations, η² (eta-squared) for ANOVA. Without effect size, statistical significance tells you almost nothing about the real-world importance of your finding. Cohen’s d and effect size analysis is the standard tool for quantifying practical significance.

Writing the Null Hypothesis for Different Statistical Tests

The structure of the null hypothesis changes slightly depending on which statistical test you are using. The core logic stays the same — no effect, no difference, no relationship — but the parameter, notation, and framing adapt to the specific test. Here is a practical reference for the most common tests encountered in college statistics courses. Selecting the right statistical test for your hypothesis is the practical decision that comes immediately after writing H0.

T-Tests: Comparing Means

One-sample t-test: Tests whether a population mean equals a specific hypothesized value. H0: μ = μ₀ (e.g., H0: μ = 100 — the population mean IQ equals 100). The one-sample t-test guide covers the full procedure. Independent samples t-test: Tests whether two group population means are equal. H0: μ1 = μ2. Paired samples t-test: Tests whether the mean difference between paired observations is zero. H0: μd = 0 (where d is the difference between paired scores). Full t-test definitions and applications break down all three variants with worked examples.

ANOVA: Comparing Multiple Group Means

Analysis of Variance (ANOVA) tests whether three or more population group means are all equal. The null hypothesis for a one-way ANOVA is: H0: μ1 = μ2 = μ3 = … = μk (all group means are equal in the population). The alternative is: Ha: At least one group mean differs from the others. ANOVA does not specify which groups differ — a post-hoc test (Tukey, Bonferroni, etc.) is needed for that. MANOVA — the multivariate extension of ANOVA — is used when you have multiple dependent variables.

Correlation Tests: Testing Relationships

When testing whether two continuous variables are related, the null hypothesis is: H0: ρ = 0 (there is no linear correlation between the variables in the population). The alternative is: Ha: ρ ≠ 0 (a correlation exists) or Ha: ρ > 0 or Ha: ρ < 0 for directional tests. Understanding correlation and statistical relationships covers Pearson’s r, Spearman’s rho, and when to use each.

Regression Analysis: Testing Predictors

In simple linear regression, the null hypothesis tests whether the regression slope (β1) equals zero — meaning the predictor variable has no linear relationship with the outcome. H0: β1 = 0 (the predictor has no effect on the outcome). Ha: β1 ≠ 0. In multiple regression, you test this for each predictor independently. The simple linear regression guide and regression analysis for predictive modeling both use this null hypothesis framework throughout their worked examples.

Chi-Square Tests: Testing Categorical Variables

For chi-square tests of independence, the null hypothesis is: H0: The two categorical variables are independent in the population (there is no association between them). For chi-square goodness-of-fit, the null is: H0: The observed frequencies match the expected distribution. The complete chi-square test guide covers both variants with full examples and worked calculations.

Essential Terms, LSI Keywords, and Concepts for Hypothesis Testing

Mastering the vocabulary of null hypothesis testing is as important as understanding the mechanics. Professors and examiners look for precise use of terminology — a student who uses these terms correctly demonstrates genuine conceptual understanding. Here is a complete reference for the core terms and related NLP concepts that appear throughout hypothesis testing coursework. Expert statistics assignment help is available when you need guidance applying these concepts to your specific assignment.

Core Hypothesis Testing Terms

Null hypothesis (H0): The default claim of no effect, no difference, or no relationship. Always includes =, ≤, or ≥. Alternative hypothesis (Ha or H1): The competing claim that an effect, difference, or relationship exists. Uses ≠, <, or >. Hypothesis test: A statistical procedure for deciding whether to reject H0 based on sample data. Test statistic: A calculated value (t, z, F, χ²) that measures how far the sample result is from the null hypothesis value. P-value: The probability of observing results as extreme as those found, assuming H0 is true. Significance level (α): The threshold for rejecting H0 — typically 0.05. Critical value: The test statistic value that marks the boundary of the rejection region. Rejection region: The range of test statistic values that lead to rejecting H0.

Type I error (α): Rejecting H0 when it is true (false positive). Type II error (β): Failing to reject H0 when it is false (false negative). Statistical power (1 − β): The probability of correctly detecting a real effect. Effect size: A measure of the magnitude of an effect, independent of sample size (e.g., Cohen’s d, r, η²). One-tailed test: A hypothesis test that specifies a directional alternative hypothesis. Two-tailed test: A hypothesis test with a non-directional alternative. Population parameter: A characteristic of a population (μ, p, ρ) — what the null hypothesis makes a claim about. Sample statistic: A characteristic of a sample (x̄, p̂, r) — calculated from data to estimate the population parameter. Expected values and variance are the mathematical foundations underlying these parameters.

Related Concepts and NLP Keywords

These related concepts frequently appear alongside null hypothesis discussions in academic literature, exam questions, and assignment prompts: research hypothesis, scientific method, inferential statistics, probability distribution, normal distribution, t-distribution, z-score, confidence interval, statistical inference, experimental design, control group, treatment group, randomized controlled trial (RCT), placebo effect, confounding variable, replication, pre-registration, open science, Bayesian statistics, frequentist statistics, NHST (null hypothesis significance testing), effect size, meta-analysis, statistical significance, practical significance, multiple comparisons, Bonferroni correction, family-wise error rate.

Many assignments ask you to contrast NHST with Bayesian approaches. Where NHST tests data against a fixed null hypothesis, Bayesian methods treat hypotheses themselves as having probability distributions — a fundamentally different philosophical approach. Both are used in modern research, and understanding the contrast enriches any statistics essay or assignment response. Decision theory provides yet another formal framework for understanding how statistical decisions are made under uncertainty, including the decision to reject or retain a null hypothesis.

How to Use These Terms in Your Assignment

The most effective assignments don’t just name these terms — they use them precisely in context. For example, instead of writing “we found our result was significant,” write: “The independent samples t-test yielded a statistically significant result (t(48) = 3.21, p = 0.002, d = 0.91), providing strong evidence to reject the null hypothesis at the α = 0.05 significance level. The large effect size (d = 0.91) indicates the finding is also practically meaningful.” This level of precision shows command of the discipline. Reporting statistical results with transparency is a skill that directly improves your academic marks and your credibility as a researcher.

Frequently Asked Questions: Null Hypothesis