Match a Research Question to Its Parameter

Start With the Quantity the Question Asks About

Before choosing a statistical procedure, identify what quantity in the population would answer the research question. Is the question about the fraction of people who experience an outcome, an average measurement, or a contrast between two groups? Choosing the parameter first keeps the analysis connected to the question, rather than to whichever calculation happens to be convenient.

A parameter describes a population. It is generally fixed for a clearly defined population, even though its value may be unknown. A statistic is calculated from sample data and can vary from sample to sample. For example, the true proportion of patients in a defined population who complete a treatment is a parameter; the proportion among 80 sampled patients is a statistic.

Definition: A population parameter is a numerical characteristic of a specified population or population distribution. For a binary outcome, \(p\) denotes a population proportion. For a quantitative outcome, \(\mu\) denotes a population mean. A sample statistic, such as \(\hat{p}\) or \(\bar{x}\), is calculated from observed sample data to describe or estimate a parameter.

A useful translation begins by stating the population, the outcome, and the requested quantity in ordinary words. Then choose notation that represents exactly that quantity. For a comparison, name both groups and settle the subtraction order before interpreting the difference. Reversing the order changes the sign, even though it concerns the same two groups.

One Population: Proportion or Mean?

A question about a proportion concerns an outcome with two defined states, such as completed or not completed. Think of each individual’s outcome as an indicator \(Y\), equal to 1 when the event of interest occurs and 0 otherwise. The population proportion \(p\) is the fraction of the population for which \(Y=1\). Equivalently, it is the population probability of that event.

$$ p=P(Y=1) $$

The event coded as 1 must match the question. If the question asks what fraction of patients experience a complication, code “complication” as the event. Asking for the fraction who do not experience one targets a different proportion.

A question about a quantitative measurement usually targets a population mean \(\mu\): the average value of that measurement in the specified population. Its units are the measurement’s units, such as days, millimetres of mercury, or minutes. If a question instead asks about a median, a particular percentile, or another feature of the distribution, a mean is not automatically the right parameter. The wording of the question determines the target.

Worked Example: A Proportion of Treatment Completions

Suppose a hypothetical rehabilitation program has 200 enrolled patients, of whom 146 complete the program. The question is, “What proportion of patients enrolled in this program complete it?” Define completion as the event of interest and the population as these 200 enrolled patients.

The target is one population proportion, \(p\), for completion. In this fully enumerated hypothetical population, its value is:

p=\frac{\text{number who complete}}{\text{number enrolled}} =\frac{146}{200}=0.730

Thus, 73.0% of this specified population completed the program. In a study that observed only a sample from a larger target population, \(p\) would still describe the larger population, but the sample fraction \(\hat{p}\) would be a statistic rather than the parameter itself. The example’s numerical calculation works because the hypothetical population has been fully counted.

Worked Example: A Mean Length of Stay

Consider a hypothetical population of four patients discharged from a particular clinic during a defined week. Their stays are 2, 3, 5, and 6 days. The question is, “What was the average length of stay for these four patients?”

Length of stay is quantitative, and the question asks for its average. The parameter for this fully specified population is its population mean, \(\mu\), measured in days:

\mu=\frac{2+3+5+6}{4} =\frac{16}{4}=4\text{ days}

The requested value is therefore a mean, not a proportion. If the intended population were all patients discharged from the clinic over a year, the same type of parameter would apply, but these four hypothetical observations alone would not give the full population mean.

Two Groups: Define the Contrast

When a research question compares two groups, first determine what is being measured in each group, then express the comparison as a difference. Use subscripts 1 and 2 to identify the groups consistently. A difference parameter is usually written as the group 1 value minus the group 2 value. Its sign and units are meaningful only when the group order and measurement have been stated.

For a binary outcome, let \(p_1\) and \(p_2\) be the proportions experiencing the event in groups 1 and 2. The difference in proportions is \(p_1-p_2\). For a quantitative outcome, let \(\mu_1\) and \(\mu_2\) be the group means; the difference in means is \(\mu_1-\mu_2\).

Definition: A difference in proportions, \(p_1-p_2\), compares the population fractions with a specified binary outcome in two named groups. A difference in means, \(\mu_1-\mu_2\), compares the population averages of a specified quantitative outcome in two named groups. The order of subtraction must be stated, and the units of a difference in means are the same as the units of the measured outcome.

These differences directly represent a comparison: \(p_1-p_2=0\) means the two population proportions are equal, while \(\mu_1-\mu_2=0\) means the two population means are equal. A positive or negative value indicates which group has the larger value under the chosen order. For a proportion difference, the units are proportion units; multiplying by 100 expresses the same difference in percentage points.

Worked Example: Comparing Two Completion Proportions

Suppose a hypothetical service has two separately defined patient groups. In group 1, 72 of 100 patients complete a program; in group 2, 61 of 100 complete it. The question asks how the population completion proportions compare, with group 1 listed first.

The target is the difference \(p_1-p_2\), where \(p_1\) is the population completion proportion for group 1 and \(p_2\) is the corresponding proportion for group 2. The given counts describe fully enumerated hypothetical groups, so their proportions and difference can be calculated:

p_1-p_2=\frac{72}{100}-\frac{61}{100} =0.72-0.61=0.11

In these hypothetical populations, the completion proportion in group 1 is 0.11 higher, or 11 percentage points higher, than in group 2. This calculation describes a contrast in proportions; it does not by itself explain why the groups differ. If the question instead asked for group 2 minus group 1, the value would be \(0.61-0.72=-0.11\), the same contrast with the reverse direction.

Worked Example: Comparing Two Mean Measurements

Imagine two hypothetical groups of patients, with recorded follow-up systolic blood pressure values in millimetres of mercury. The fully enumerated values are 124, 130, and 136 for group 1, and 128, 132, and 134 for group 2. The question is, “What is the difference in average follow-up blood pressure, group 1 minus group 2?”

The outcome is quantitative, so the target is a difference in population means, \(\mu_1-\mu_2\), measured in millimetres of mercury. For these fully enumerated hypothetical groups:

\mu_1-\mu_2 =\frac{124+130+136}{3}-\frac{128+132+134}{3} =130-131.33=-1.33\text{ mm Hg}

The mean in group 1 is 1.33 mm Hg lower than the mean in group 2 for these specified groups. The answer is not a proportion difference: the outcome is a numerical measurement, and the research question asks to compare averages.

When the Same Individuals Are Measured Twice

A comparison can also involve two measurements on each individual, such as a value before and after an intervention. The parameter should then reflect the within-individual comparison. Choose a direction, define each person’s difference in that direction, and target the population mean of those differences. This is not the same definition as subtracting the means of two unrelated groups.

For example, define \(D\) as post-treatment measurement minus pre-treatment measurement. Then \(\mu_D\) denotes the population mean change under that definition. A positive value represents an average increase, and a negative value an average decrease. The unit is the unit of the original measurement.

Worked Example: Average Change in Paired Measurements

In a hypothetical set of three people, a measurement is recorded before and after a program. The before values are 10, 12, and 15 units; the after values are 13, 11, and 17 units. The question asks for the average within-person change, defined as after minus before, for these three people.

Compute a difference for each person using the stated order: \(D_1=13-10=3\), \(D_2=11-12=-1\), and \(D_3=17-15=2\) units. The target for this fully enumerated hypothetical population is the mean paired difference:

\mu_D=\frac{3+(-1)+2}{3} =\frac{4}{3}=1.33\text{ units}

The average change is an increase of 1.33 units under the after-minus-before definition. Reversing the order would make every individual difference, and the mean difference, negative. Stating the order therefore matters even when the measurements themselves are unchanged.

Why the Difference Matches the Question

A difference parameter is not merely a convenient formula: it states the exact contrast being asked about. For two groups, if the group means are \(\mu_1\) and \(\mu_2\), their difference in the chosen order is \(\mu_1-\mu_2\). Reversing the order gives \(\mu_2-\mu_1\), which is the negative of the first contrast:

\mu_2-\mu_1=-(\mu_1-\mu_2)

The same relationship holds for proportions. For paired measurements, defining \(D\) as the individual change and averaging those changes targets the average within-person change. These definitions explain why naming groups, outcomes, and subtraction order is part of identifying the parameter—not a detail to fill in after a calculation.

A careful parameter statement can be made in one sentence: “The target is the difference in population mean follow-up blood pressure, in mm Hg, between group 1 and group 2, defined as group 1 minus group 2.” That sentence identifies the quantity, population groups, units, and direction without claiming that any particular procedure has been selected.

Common Mistakes and a Practical Check

Confusing a statistic with a parameter. The proportion in a sample, \(\hat{p}\), is not the population proportion \(p\). State whether the question concerns the observed sample or a broader population.
Choosing a proportion for any question involving a percentage. A percentage might be a way of reporting a difference in proportions, but a question about average blood pressure targets a mean, not a proportion.
Leaving the comparison direction unstated. “Difference between the groups” does not specify a sign. Write group 1 minus group 2, and keep that order in the interpretation.
Treating paired measurements as separate group averages. If the question asks about average within-person change, define each person’s difference first and target its population mean.
Claiming a cause from a contrast alone. A difference parameter describes how two population quantities compare. The parameter’s definition by itself does not establish that group membership caused the difference.

Before moving from a question to any calculation, check four things: the population or groups are defined; the outcome is identified; the target is a proportion, mean, or difference that matches the wording; and the direction and units are clear. This check narrows the statistical target without yet deciding how to estimate it or assess uncertainty.

Check Your Understanding

For each question, identify the target parameter in words and notation. State a group order where a difference is involved.

A clinic asks what fraction of its patients miss at least one scheduled appointment during a year. What population quantity is targeted?
A researcher asks for the average number of days between diagnosis and treatment among patients in a defined region. Identify the parameter and its units.
A report compares the fraction of two clinics’ patients who return for follow-up. Define the difference parameter and explain why the clinic order matters.
A study asks whether average recovery time differs between two named patient groups. State the parameter for group 1 minus group 2 and its units.
For each participant, a measurement is recorded before and after an exercise program. If change is defined as after minus before, what parameter describes the average within-person change?