A hypothesis is a testable prediction about a relationship between variables, stated precisely enough that data can contradict it. Vague hypotheses cannot be tested, and hypotheses that merely restate the research question add nothing.
Enter what you want to test and this tool drafts matched null and alternative pairs for each of the main relationship types, naming the statistical test that goes with each. It also covers the reporting conventions that trip students up in the results chapter.
The null hypothesis states there is no difference, relationship or effect. The alternative states there is one. You test the null, because it is the specific claim that data can contradict — statistical tests calculate how likely your data would be if the null were true.
This is why the wording of your conclusion matters. You either reject the null or fail to reject it. You never prove the alternative, and you never prove the null: failing to reject may only mean your sample was too small to detect an effect that exists.
A difference between two groups calls for an independent-samples t-test; three or more groups, a one-way ANOVA. A relationship between two continuous variables calls for Pearson correlation, or Spearman if the data is ordinal or badly skewed. An effect of one variable on another calls for regression.
An association between two categorical variables calls for a chi-square test of independence. Choosing the test before you collect data is not premature — it determines what you must measure and at what level, and discovering afterwards that your data cannot support the intended test is a common and avoidable disaster.
Only if your study is quantitative and testing a specific prediction. Exploratory and qualitative studies use research questions instead, and forcing hypotheses onto interview-based work is a category error that examiners notice immediately.
Descriptive quantitative work sits in between: if you are simply establishing how common something is, a research question serves better than a hypothesis. Hypotheses belong where you are comparing, correlating or testing an effect.
A two-tailed hypothesis says there is a difference without specifying direction, and is the default. A one-tailed hypothesis predicts direction — that group A will score higher — and concentrates all the statistical power on that side.
That makes one-tailed tests easier to reach significance with, which is exactly why they need justification from prior literature rather than convenience. Choosing one-tailed after seeing which way your data leans is a serious methodological error, and it is detectable.
State your significance level in advance, conventionally 0.05. Report the test statistic, the degrees of freedom, the p-value and an effect size. A p-value tells you whether an effect is distinguishable from zero; it says nothing about how large or important it is.
With a large enough sample, trivial differences become statistically significant. Effect size — Cohen's d, eta squared, r — is what tells your reader whether the finding matters in practice, and increasingly examiners expect it as standard rather than as an optional extra.
The null states there is no difference, relationship or effect; the alternative states there is one. You test the null, since it is the claim data can contradict.
No. Qualitative and exploratory studies use research questions. Hypotheses belong to quantitative work testing specific predictions.
No. You reject or fail to reject the null. Failing to reject is not evidence the null is true — it may simply mean insufficient statistical power.
Usually one per objective. More than four or five in a Master's thesis normally indicates the scope is too broad.
0.05 is conventional in most social science. Choose it before analysis, never after seeing your results.