Doctors Revision

Measures of Central Tendency: Mean, Median and Mode

Text size
On this page 20 sections

Overview

Clinical Medicine Year 2 • Descriptive biostatistics

Measures of central tendency summarise where observations tend to cluster. The arithmetic mean, median and mode each describe a different aspect of “typical,” so no one measure is automatically correct for every health variable. A good summary depends on the measurement scale, distribution, outliers, study question and audience. In clinical and public-health reports, a measure of centre should usually be shown with a measure of spread and a clear denominator, not presented alone.

Learning outcomes

By the end, you should be able to calculate the mean, median and mode; choose among them using scale and distribution; interpret a frequency distribution; explain how skew and outliers affect a summary; and report centre together with appropriate spread in health research.

1. What does “central” mean?

A measure of central tendency is a numerical summary of the location or typical value of a distribution. It condenses many observations into a value that is easier to compare, but it does not describe all features of the data. Two clinics could have the same mean waiting time while one has tightly grouped values and the other has a long tail of very delayed patients. Always inspect the distribution and consider spread, shape and context.

Before calculating, identify the variable and its scale. Nominal categories have no natural order; ordinal categories do; continuous measurements can take values along a scale. A mean of blood group codes is meaningless, even if categories are stored as numbers. A mean of a carefully measured interval-scale value may be useful when interpreted with the data distribution.

2. Arithmetic mean

The arithmetic mean is the sum of observed values divided by the number of valid observations. For a sample, it is written as x̄ = Σxᵢ/n; the population mean is commonly written μ. The mean uses every observation and has useful mathematical properties, which is why it appears in many statistical methods.

Strengths of the mean

  • Uses all valid numerical observations, rather than only their order or most frequent value.
  • Supports further calculations such as variance, standard deviation and confidence intervals under suitable assumptions.
  • Is often informative for reasonably symmetric distributions without influential extreme values.
  • Can be combined across groups if group sizes and means are known: the overall mean is a size-weighted mean of group means.

Limitations of the mean

The mean can be pulled toward a long tail or extreme value. It can be non-representative for skewed variables such as household expenditure, hospital length of stay or waiting time. It is inappropriate for purely nominal categories and may be difficult to interpret for an ordinal scale where category spacing is not known to be equal.

3. Median

The median is the middle value after observations are ordered from smallest to largest. With an odd number of observations, it is the single middle value. With an even number, it is commonly the average of the two middle values. Half the observations are at or below the median and half are at or above it, subject to ties.

The median is resistant to extreme values because it depends primarily on rank rather than the magnitude of every observation. This makes it useful for skewed continuous data and ordered outcomes. It does not use the full distance between observations, so it may be less informative when the distribution is symmetric and the mean is a meaningful summary. A median also needs a meaningful order; it cannot summarise nominal categories such as blood group.

4. Mode

The mode is the value or category observed most often. It is the only one of these three measures that applies directly to nominal variables. For example, the most frequently recorded blood group or diagnosis category can be described using the mode. A dataset can have one mode (unimodal), two modes (bimodal), multiple modes or no uniquely most frequent value. For continuous data, a mode can depend on measurement precision or how values are grouped into bins, so a histogram may be more useful than a single exact value.

Mode is useful for describing the commonest response or category, but it may ignore much of the distribution. If two categories occur almost equally often, reporting only one “most common” category may hide the near tie. State frequencies or percentages as well.

5. Worked health example: clinic waiting time

Illustrative values only—not facility data: imagine eight patients have recorded waits of 10, 12, 12, 14, 16, 18, 20 and 95 minutes. The sum is 197 minutes, so the mean is 197/8 = 24.625 minutes (about 24.6). The two middle values are 14 and 16, so the median is 15 minutes. The mode is 12 minutes because it appears twice and all other values appear once.

The mean is much higher than most individual waiting times because the 95-minute observation pulls it upward. That value might be a real delay, an unusual case, a recording problem or a data-entry error; investigate it rather than deleting it automatically. The median better reflects a typical patient in this right-skewed example, while the maximum and distribution are also important for service planning. A fuller report could provide median and interquartile range plus the range or a high percentile when extreme waits matter.

6. Choosing a measure for the data

Data and pattern Useful centre Why / what to add
Data and patternUseful centreWhy / what to add
Nominal categories, such as diagnosis groupMode, counts and percentagesThere is no meaningful numeric order; report denominators and missing values.
Ordinal rating or ordered stageMedian and category frequencies; sometimes modeOrder is meaningful but spacing between categories may not be equal.
Continuous, roughly symmetric without extreme outliersMeanOften report standard deviation to describe spread; inspect a plot.
Continuous and strongly skewed or with influential extremesMedianOften report interquartile range; show distribution or clinically important percentiles.
Two distinct subgroups or a multimodal patternConsider stratified summaries, not one pooled centreA single mean or median can describe neither subgroup well.

This is a guide rather than a rigid rule. The analysis plan, clinical meaning and sampling design matter. For a skewed cost or length-of-stay outcome, a mean may still be needed for resource planning because total cost depends on it; report it with a robust summary and explain the purpose.

7. Frequency tables and grouped observations

For a frequency table, record each distinct value or category and its frequency. For discrete data, the mode is the value with the largest frequency. A grouped-data mean can be approximated by multiplying each class midpoint by its frequency, summing those products, and dividing by the total frequency. This is only an approximation because individual observations within a class are unknown.

Number of clinic visits missed in six months Patients (frequency) Contribution to total
Number of clinic visits missed in six monthsPatients (frequency)Contribution to total
0400 × 40 = 0
1251 × 25 = 25
2102 × 10 = 20
Total7545 missed visits

The mean number missed per patient is 45/75 = 0.6 visits. The median is 0 because at least 40 of the 75 patients have zero missed visits, placing the middle observation at zero. The mode is also 0. Each summary answers a different question: average burden, middle patient and most common value.

8. Mean, median and mode in distribution shape

In a perfectly symmetric, unimodal distribution, the mean, median and mode may coincide or be close. In a right-skewed distribution, a few high observations often pull the mean above the median. In a left-skewed distribution, low values may pull the mean below the median. These relationships are useful clues, not a substitute for a histogram, box plot or subject-matter interpretation.

A bimodal distribution may indicate two subpopulations—for example, two clinics with different staffing patterns or two patient groups with different severity. Pooling them can produce a “centre” that is uncommon in either subgroup. Stratify by meaningful, prespecified groups and explain why the groups differ.

9. Centre should be paired with spread

Central tendency alone hides variability. For reasonably symmetric numerical data, report mean with standard deviation (SD); SD describes spread of observations, not uncertainty in the mean. For skewed data, report median with interquartile range (IQR), where the IQR spans the middle half of observations. Range and percentiles may add useful information, especially for safety or service thresholds.

Do not confuse standard deviation with standard error. SD describes variation among observed values; standard error describes uncertainty in an estimated statistic under a sampling model. For comparisons or population inference, a confidence interval addresses uncertainty in the estimate. See confidence intervals in health research.

10. Missing values and outliers

Calculate summaries using the correct valid denominator. Missing is not zero, and a code such as 99 can distort the mean if it is not treated according to the data dictionary. Report how many observations contribute to each summary and whether missingness differs by group. Do not silently exclude participants or treat an incomplete record as a negative response.

An outlier may be a data-entry error, a measurement error, a rare but valid observation or a sign that the sample contains different subgroups. Check the source, units and protocol. If the value is valid, consider whether the mean is appropriate, show a robust summary or perform a planned sensitivity analysis. Deleting a value solely because it changes the conclusion is not acceptable.

11. Applying the measures in clinical and public-health practice

12. Common mistakes

13. Mini case and review questions

A district reports median time from arrival to clinician assessment of 18 minutes at Facility A and 18 minutes at Facility B. Does this prove that service performance is identical? No. The medians may match while one facility has many patients seen quickly and a few very long waits. Compare distributions, IQRs and relevant upper percentiles; review case mix, staffing and measurement. If the goal is overall workload, means may also be needed.

  1. Calculate the mean, median and mode for 3, 4, 4, 7 and 12.
  2. Which measure is least affected by one extremely high value?
  3. Which measure is appropriate for nominal categories?
  4. Why might reporting mean and SD alone be misleading for skewed length of stay?
  5. What does the IQR describe?

Continue learning

Review presentation of data in descriptive biostatistics, hypothesis testing, confidence intervals and data-analysis workflows in SPSS, STATA and Excel. Return to the Clinical Medicine Year 2 curriculum.

References and further reading

Educational revision content only. Select and interpret summaries in light of the measurement scale, study question, data quality and distribution. Use appropriate statistical guidance for inferential analysis.

Further Reading

For deeper study, use authoritative biostatistics and reporting resources:

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top