From a Verbal Claim to a Statistical Question
A statistical claim often arrives in ordinary language: a clinic says that its average waiting time is 20 minutes, a manufacturer claims that at least 95% of its products meet a standard, or a community group suspects that a population proportion has changed. Before data can address such a claim, it must be expressed in terms of a population parameter. The direction of the claim matters: “greater than,” “less than,” and “different from” lead to different alternative hypotheses.
This translation is a matter of defining what is being described and what departure from a benchmark is of interest. It does not, by itself, say whether the claim is supported by data. That judgment requires an appropriate statistical test and its results.
The parameter must describe the population relevant to the question. For example, if the question concerns all patients treated at a particular clinic during a defined period, \(\mu\) might be their population mean waiting time, measured in minutes. If the question concerns the fraction of all devices produced by a factory that pass inspection, \(p\) is that population proportion. Defining the population and variable prevents a mathematically correct pair of hypotheses from answering the wrong question.
Identify the Parameter Before Choosing a Direction
A mean claim concerns a quantitative measurement and is written using \(\mu\). A proportion claim concerns the fraction of individuals in a specified population who have a particular characteristic and is written using \(p\). The benchmark is the value used to describe the stated standard or comparison. It is a fixed reference value, not a quantity calculated from the sample.
The sample mean \(\bar{x}\) and the sample proportion \(\hat{p}\) are statistics: they summarize observed sample data. They are not the parameters in the hypotheses. The hypotheses describe the population, even though sample data may later be used to assess them.
| Verbal claim or question | Parameter | Alternative direction |
|---|---|---|
| The population mean is greater than a benchmark | \(\mu\) | \(H_a:\mu>\mu_0\) |
| The population mean is less than a benchmark | \(\mu\) | \(H_a:\mu<\mu_0\) |
| The population mean differs from a benchmark | \(\mu\) | \(H_a:\mu\ne\mu_0\) |
| The population proportion is greater than a benchmark | \(p\) | \(H_a:p>p_0\) |
| The population proportion is less than a benchmark | \(p\) | \(H_a:p<p_0\) |
| The population proportion differs from a benchmark | \(p\) | \(H_a:p\ne p_0\) |
Here \(\mu_0\) denotes a benchmark for a population mean, and \(p_0\) denotes a benchmark for a population proportion. The symbols \(>\) and \(<\) retain their usual meanings: the parameter is being claimed to be above or below the benchmark. “Different” means that departures in either direction matter, so its alternative uses \(\ne\).
In the standard hypothesis-pair form used here, the null hypothesis places the parameter at the benchmark: \(H_0:\mu=\mu_0\) or \(H_0:p=p_0\). The alternative expresses the direction of interest. Writing the benchmark explicitly makes it easier to see which population quantity is under discussion and what comparison the claim makes.
Worked Examples: Translating Claims
Worked Example: A Claim About a Greater Mean
A hypothetical transit agency wants to investigate whether the mean weekday bus delay for all buses on a specified route is greater than 6 minutes. Define the parameter and write the hypotheses.
Step 1: Identify the population and measurement. The population is all weekday bus trips on the specified route during the period of interest. The quantitative measurement is delay in minutes.
Step 2: Name the parameter. Let \(\mu\) be the population mean delay, in minutes, for those trips. The benchmark is \(\mu_0=6\) minutes.
Step 3: Translate the direction. The phrase “greater than 6 minutes” says \(\mu>6\), so that is the alternative. The null places the mean at the stated benchmark:
The units help preserve the meaning of the parameter. The statement is about a population average delay, not about whether one bus is delayed by more than 6 minutes and not about whether the sample mean \(\bar{x}\) is above 6.
Worked Example: A Claim About a Different Mean
In a hypothetical laboratory quality review, investigators ask whether the population mean fill volume of bottles differs from the labeled value of 500 milliliters. Translate the question into hypotheses.
Step 1: Identify the parameter. Let \(\mu\) be the mean fill volume, in milliliters, for all bottles produced under the process being reviewed. The comparison value is \(\mu_0=500\) milliliters.
Step 2: Read the wording carefully. “Differs from 500 milliliters” does not specify whether the mean is higher or lower. Both directions are relevant, so the alternative is two-sided:
For instance, a mean of 498 mL and a mean of 502 mL are both departures in the direction of the alternative. Choosing \(H_a:\mu>500\) would instead ask only whether the process overfills on average; it would not represent the stated question about any difference.
Worked Example: A Claim About a Greater Proportion
A hypothetical consumer group wants to investigate whether more than 8% of repairs made by a service company require a repeat visit within 30 days. Define the population proportion and write the hypotheses.
Step 1: Define the population and outcome. The population is all repairs made by the company under the conditions covered by the investigation. The outcome is whether a repair requires a repeat visit within 30 days.
Step 2: Name the parameter. Let \(p\) be the proportion of those repairs that require a repeat visit within 30 days. The benchmark is \(p_0=0.08\), or 8%.
Step 3: Translate the claim. “More than 8%” means the population proportion is greater than 0.08. Thus:
This is a claim about a proportion, not a mean. The statistic \(\hat{p}\), calculated from a sample of repairs, would describe that sample; \(p\) in the hypotheses describes the population of repairs.
Worked Example: An “At Least” Claim Being Challenged
A hypothetical producer states that at least 92% of its packages meet a weight standard. A regulator wants to investigate whether the true proportion is below 92%. Translate the producer’s claim and the regulator’s question into a hypothesis pair.
Step 1: Define the parameter. Let \(p\) be the proportion of all packages made under the process of interest that meet the weight standard. The benchmark is \(p_0=0.92\).
Step 2: Separate the original claim from the question being tested. The producer’s words “at least 92%” express \(p\ge 0.92\). The regulator’s concern is specifically that the proportion may be below the benchmark, which is the direction \(p<0.92\).
Step 3: Write the standard benchmark form of the hypotheses. The null uses the benchmark and the alternative represents the regulator’s downward concern:
This pair makes the direction of the investigation clear. It does not mean that the producer’s literal wording was “exactly 92%”: the producer asserted an inclusive lower bound, while the regulator is investigating a shortfall from that bound. Keeping the verbal claim distinct from the benchmark form avoids changing the meaning of the original statement.
Why the Direction Follows From the Question
A useful way to justify the translation is to write down which parameter values the question treats as departures of interest. Suppose a benchmark is \(\theta_0\), where \(\theta\) stands for either \(\mu\) or \(p\). If the question asks whether the parameter is higher, the values of interest satisfy \(\theta>\theta_0\). If it asks whether the parameter is lower, those values satisfy \(\theta<\theta_0\). If any departure matters, the values of interest satisfy \(\theta\ne\theta_0\).
The null and alternative should refer to the same parameter and benchmark, with the alternative expressing precisely the departure being investigated. This can be checked directly from the inequalities:
The first two alternatives point in one direction; the third includes both. This is why the phrase “has changed” often calls for a two-sided alternative, while “has increased” calls for an upper-direction alternative. Replacing “increased” with “changed” is not a harmless wording change: it changes which departures the hypotheses represent.
State who or what the claim describes and, for a proportion, what counts as having the characteristic.
Use \(\mu\) for a population mean or \(p\) for a population proportion, and record the numerical comparison value.
Use \(>\) for greater, \(<\) for less, and \(\ne\) when departures in either direction matter.
Put the benchmark in \(H_0\), write the investigative direction in \(H_a\), and confirm both concern the same population parameter.
Common Mistakes and Why They Matter
- Putting a sample statistic in the hypotheses. Writing \(H_a:\bar{x}>6\) describes a sample result, not a population claim. Write \(H_a:\mu>6\); \(\bar{x}\) is a statistic that may later provide evidence about \(\mu\).
- Choosing the direction from what seems plausible. The wording “different from 500” does not justify a greater-than alternative just because overfilling seems more likely. Use \(H_a:\mu\ne500\) unless the question specifically limits interest to one direction.
- Confusing a mean with a proportion. A statement about the fraction of repairs needing another visit concerns \(p\), even if the data are recorded as Yes/No values. A statement about average repair time concerns \(\mu\).
- Dropping the context or units. “\(\mu=6\)” is incomplete if the reader cannot tell which population mean is meant. Define \(\mu\) and include the measurement and units in the surrounding explanation.
- Assuming every verbal claim becomes the alternative. A claim may be the benchmark being challenged, as in “at least 92%,” while the investigation asks whether the population falls below it. State both the original claim and the direction under investigation before writing the pair.
Check Your Understanding
For each item, identify the parameter and benchmark, then write an appropriate null and alternative hypothesis.
- A health service asks whether the population mean appointment wait is less than 25 minutes. What parameter and direction belong in the hypotheses?
- A company investigates whether the proportion of its products that fail inspection differs from 3%. Is the alternative one-sided or two-sided?
- A report claims that at least 80% of eligible residents received a service. A researcher investigates whether the proportion is below 80%. Distinguish the report’s claim from the alternative direction.
- Explain why \(H_a:\hat{p}>0.10\) is not an appropriate hypothesis about a population proportion.
- A school wants to know whether the mean score differs from 70 points, without specifying higher or lower. What symbol belongs in the alternative, and why?