Descriptive Statistics and Data Visualization

Introduction and Context

Human resource managers make decisions about recruitment, absence, performance, compensation, training, and employee experience. These decisions depend on data, but raw data are difficult to interpret when presented as long lists of observations. Descriptive statistics converts collected data into meaningful summaries, while data visualization communicates patterns, comparisons, and unusual observations efficiently.

Descriptive statistics does not prove that one variable causes another. It describes the observed dataset. Statistical conclusions about a wider population require inferential methods, which are studied separately. Nevertheless, accurate descriptive analysis is the foundation of sound inference and responsible managerial decision-making.

Learning Objectives

By the end of this chapter, students should be able to:

1. Distinguish populations, samples, variables, and measurement scales.
2. Construct and interpret frequency distributions and cross-tabulations.
3. Calculate and evaluate measures of central tendency, dispersion, position, and shape.
4. Select suitable charts for different types of HR data.
5. Detect misleading or ethically problematic visualizations.
6. Apply descriptive statistics to a practical HR case.

1. Data, Variables, and Measurement

A population is the complete set of units relevant to a study, such as all employees in an organisation. A sample is a subset selected from that population. An observation is one recorded unit, and a variable is a characteristic that can take different values.

Variables may be:

• Categorical: department, employment type, or preferred benefit.
• Numerical: age, salary, training hours, or absence days.
• Discrete: countable values, such as number of absences.
• Continuous: values on a continuum, such as time or weight.

Measurement scales determine which summaries are meaningful.

A parameter describes a population; a statistic describes a sample. For a finite population of N values, the population mean is μ. For a sample of n values, the sample mean is x̄. Careful analysts identify the unit of analysis, time period, source, missing-value treatment, and definitions before calculating anything.

2. Organising Data: Frequency Distributions

A frequency distribution records how often each value or class occurs. For category i, frequency is fᵢ and total frequency is n = ∑fᵢ. Relative frequency is:

Percentage frequency is relative frequency × 100. For numerical data with many distinct values, intervals should be mutually exclusive, collectively exhaustive, and sufficiently informative. Class width is often approximated by:

The result is rounded upward to a practical value. This is a planning rule, not a mathematical law.

Worked construction: absence days

Suppose 20 employees recorded the following absence days in one quarter:

2, 0, 1, 3, 2, 5, 1, 0, 4, 2, 1, 6, 3, 2, 0, 1, 4, 2, 3, 1.

Grouping into 0–1, 2–3, 4–5, and 6–7 days gives:

The distribution shows that 80% of observations fall between 0 and 3 days. Grouping simplifies communication but loses detail; therefore, the original data should be retained.

3. Measures of Central Tendency

Mean

For n observations x₁, x₂, …, xₙ, the arithmetic sample mean is:

The mean uses every observation and is useful for numerical ratio data. It is sensitive to extreme values. For frequency data:

Here xᵢ is a value, fᵢ its frequency, and n = ∑fᵢ. For grouped classes, the class midpoint is used as xᵢ, so the result is an estimate rather than an exact mean.

Median

The median is the middle ordered value. If n is odd, its position is (n + 1) ÷ 2. If n is even, it is the average of the values in positions n ÷ 2 and (n ÷ 2) + 1. The median is resistant to outliers and is often preferable for salary data.

Mode

The mode is the most frequent value or category. A dataset may have one mode, several modes, or no repeated value. It is especially useful for nominal variables, where a mean is meaningless.

Worked comparison

For training hours 3, 5, 6, 8, and 13:

The ordered middle value is 6 hours, so the median is 6. The mean exceeds the median because the value 13 pulls the mean upward. There is no mode because every value occurs once.

A weighted mean is appropriate when observations have different importance:

where wᵢ is a non-negative weight. Weights must be defined consistently; an unexamined weighting scheme can distort results.

4. Measures of Dispersion

Two teams can have the same average performance but very different consistency. Dispersion measures quantify spread.

Range and interquartile range

The range is simple but depends only on two observations. The interquartile range is:

where Q₁ is the 25th percentile and Q₃ is the 75th percentile. The IQR describes the middle half of ordered observations and is less affected by extremes.

Variance and standard deviation

For a sample:

The denominator n − 1 is used when estimating population variability from a sample. For a complete population, the population variance uses N rather than N − 1:

Variance is expressed in squared units, whereas standard deviation returns to the original units. A standard deviation is not automatically a measure of error or poor performance; its meaning depends on the variable and context.

Worked calculation

For the sample 2, 4, and 6, x̄ = 4. Deviations are −2, 0, and 2; squared deviations total 8.

Thus the sample standard deviation is 2 units.

Coefficient of variation

For positive ratio-scale data, the coefficient of variation compares relative dispersion:

It should not be used when the mean is zero or close to zero, or for arbitrary-origin scales. It is useful, for example, when comparing variability in training hours and overtime hours, provided the means are positive and interpretation is meaningful.

5. Measures of Position and Shape

A percentile is a value below which a specified percentage of observations lies. Quartiles divide ordered data into four parts. Percentiles require a stated calculation convention because software packages may interpolate differently; reports should identify the method when precision matters.

The standardized score is:

for sample-based standardization, or z = (x − μ) ÷ σ when population parameters are known. It indicates how many standard deviations an observation lies above or below the reference mean.

A distribution is symmetric when its two sides are broadly balanced. Right skewness has a longer upper tail and often mean > median; left skewness has a longer lower tail and often mean < median. These relationships are tendencies, not universal rules. Skewness statistics can be computed, but visual inspection and substantive knowledge remain important.

The empirical rule states that for an approximately normal distribution, about 68%, 95%, and 99.7% of observations lie within 1, 2, and 3 standard deviations of the mean, respectively. This rule is not valid merely because a dataset has a mean and standard deviation; approximate normality is required.

6. Data Visualization

A good visualization has a clear purpose, accurate scales, readable labels, an informative title, a source, and an identified unit of analysis. Chart selection follows the data structure.

Worked bar chart

A survey of 40 employees recorded preferred learning format:

Text-based chart, where each block represents one employee:

Classroom          ████████████ 12
Online self-paced  ████████████████ 16
Blended            ████████ 8
Coaching           ████ 4

The online self-paced format is the modal category. The chart should not imply that preferences are continuous numerical measurements, and percentages should use the same denominator of 40.

Worked time-series chart

Monthly voluntary turnover rates were January 4%, February 5%, March 4%, and April 7%. A line chart would place months on the horizontal axis and percentages on the vertical axis, connecting the four observations in chronological order. The increase from March to April is 3 percentage points, not “a 3% increase.” Relative increase is:

Both statements are mathematically correct but communicate different quantities.

Worked cross-tabulation

The following table classifies 100 employees by work arrangement and whether they completed required compliance training.

Row percentages are calculated using each row total. Completion rates are therefore 42 ÷ 50 = 84% on-site, 30 ÷ 40 = 75% hybrid, and 7 ÷ 10 = 70% remote. The denominators matter: comparing counts alone would be misleading because groups differ in size. A grouped bar chart could display completion and non-completion percentages for each arrangement. This table describes association in the sample; it does not establish that work arrangement causes completion.

Visual integrity

Misleading practices include truncating a bar-chart axis to exaggerate differences, using unequal interval widths, changing denominators without notice, using three-dimensional effects, hiding missing data, and presenting a correlation as causal evidence. Colour should not be the only way to distinguish categories, because accessibility requires adequate contrast and alternative labels.

7. Descriptive Analysis Workflow

A defensible workflow is:

1. Define the managerial question and population.
2. Identify the unit of analysis and variable types.
3. Inspect missing values, duplicates, impossible values, and outliers.
4. Produce counts and percentages for categorical variables.
5. Produce appropriate centre and spread measures for numerical variables.
6. Visualise distributions and relevant comparisons.
7. Compare groups using consistent denominators and definitions.
8. Document transformations, exclusions, software settings, and dates.
9. Interpret patterns cautiously and communicate limitations.

Software such as spreadsheets, statistical packages, and HR information systems can calculate summaries rapidly. Automation does not replace validation: analysts must check formulas, filters, coding, data access, and whether charts reflect the underlying records. AI tools may assist with chart suggestions or anomaly screening, but confidential employee data should not be uploaded to unauthorised systems, and generated interpretations require human review.

8. Illustrative HR Case: Absence and Work Arrangement

Background and problem

An organisation wants to understand quarterly absence among 80 employees in on-site, hybrid, and remote roles. Management asks whether one arrangement should be changed.

Analysis

The analyst first checks that absence days are measured over the same 90-day period and that employees who joined partway through the quarter are identified. A frequency distribution describes the overall pattern. A cross-tabulation compares work arrangement with an indicator such as “at least one absence day,” while medians and IQRs compare absence days because the distribution is likely to be skewed.

The analyst reports group sizes, missing records, medians, IQRs, and a box plot rather than relying only on a single overall mean. If a small remote group has a high average because of one unusual case, the median and individual review provide essential context. The analysis does not infer that arrangement caused absence, because job type, health conditions, commuting demands, manager practices, and selection into work arrangements may confound the comparison.

Approach and possible outcome

Management uses the results to investigate workload, flexibility, and attendance-recording practices rather than imposing an immediate policy change. A follow-up study with a defined observation period and relevant covariates can examine the issue more rigorously.

Learning points

• A descriptive comparison can identify patterns without proving causes.
• Denominators, missing data, and group sizes affect interpretation.
• Privacy protection is essential when absence data may reveal health information.
• Responsible HR decisions combine statistical summaries with organisational context.

9. Limitations, Ethics, and Governance

Descriptive statistics can conceal subgroup differences, measurement error, and changing definitions. Averages may obscure inequality; a chart may omit people with missing records; and an outlier may be either an error or an important employee experience. Analysts should preserve an audit trail, use access controls, minimise personally identifiable information, and report only what is necessary for the decision.

HR data may be sensitive under privacy and employment regulations. Organisations should establish lawful purpose, transparency, retention limits, role-based access, and procedures for correcting inaccurate records. Fairness requires checking whether a summary or visualisation disproportionately hides or exposes a group. Small-cell reporting can create re-identification risk, especially when demographic categories are combined. Statistical summaries should support, not replace, procedural fairness and meaningful human oversight.

Chapter Summary

Descriptive statistics organises and summarises observed data without automatically making causal or population-wide claims. Sound analysis begins by identifying populations, samples, variables, measurement scales, units, and data-quality problems. Frequency distributions, cross-tabulations, means, medians, modes, ranges, IQRs, variances, standard deviations, percentiles, and standardised scores answer different questions and have different assumptions. Visualisations such as bar charts, histograms, line charts, scatter plots, box plots, and heatmaps should be selected according to data type and analytical purpose. Accurate denominators, transparent scales, privacy safeguards, and cautious interpretation are essential in HR analytics. Technology can improve speed, but professional judgement and governance remain indispensable.

Key Terms

Population: The complete set of units relevant to a study.

Sample: A subset of a population.

Variable: A characteristic that can take different values.

Observation: One recorded unit in a dataset.

Parameter: A numerical description of a population.

Statistic: A numerical description of a sample.

Frequency: The number of times a value or category occurs.

Relative frequency: Frequency divided by total observations.

Mean: The arithmetic average.

Median: The middle ordered value.

Mode: The most frequent value or category.

Range: Maximum minus minimum.

Quartile: A value dividing ordered data into four sections.

Interquartile range: Q₃ minus Q₁.

Variance: Average squared deviation, using a specified population or sample denominator.

Standard deviation: The square root of variance.

Percentile: A value below which a stated proportion of observations lies.

Histogram: A chart of numerical distributions using adjacent intervals.

Cross-tabulation: A table showing joint frequencies for categorical variables.

Outlier: An unusually distant observation relative to the data pattern.

Review Questions

Short Answer Questions

1. Distinguish a population parameter from a sample statistic.
2. Why is the median often useful for analysing salaries?
3. Define the interquartile range and state what it measures.
4. What is the difference between a bar chart and a histogram?
5. Why must denominators be reported when comparing HR groups?

Long Answer Questions

1. Explain the four measurement scales and give an HR example of each.
2. Discuss the strengths and limitations of the mean, median, and mode.
3. Explain sample variance, sample standard deviation, and the role of n − 1.
4. Develop guidelines for selecting visualisations for categorical, numerical, and time-series data.
5. Discuss ethical and governance requirements for descriptive HR analytics.

Analytical / Application Questions

1. A department records salaries of 28, 30, 31, 32, and 79 thousand currency units. Calculate the mean and median, identify the likely effect of the high value, and recommend a suitable summary for a typical employee.
2. Construct a frequency distribution for the absence data in this chapter and explain what information is lost through grouping.
3. Using the work-arrangement cross-tabulation, calculate each group’s completion rate and explain why the results should not be interpreted as causal evidence.

Multiple Choice Questions with Answers

Question 1

Which measure is most appropriate for a nominal variable such as department?

A. Mean

B. Standard deviation

C. Mode

D. Interquartile range

Correct Answer: C. Mode

Question 2

For the values 3, 5, 6, 8, and 13, what is the arithmetic mean?

A. 6

B. 6.5

C. 7

D. 8

Correct Answer: C. 7

Question 3

Which measure is least affected by an extreme outlier?

A. Mean

B. Median

C. Variance

D. Range

Correct Answer: B. Median

Question 4

If a sample has 20 observations and a category occurs 5 times, its relative frequency is:

A. 0.05

B. 0.20

C. 0.25

D. 4.00

Correct Answer: C. 0.25

Question 5

For a sample of 2, 4, and 6, the sample standard deviation is:

A. 1

B. 2

C. 4

D. √8

Correct Answer: B. 2

Question 6

Which display is generally most suitable for showing the distribution of a continuous numerical variable?

A. Histogram

B. Pie chart

C. Bar chart with separated categories

D. Organisational chart

Correct Answer: A. Histogram

Question 7

In a cross-tabulation, a row percentage is calculated using:

A. The grand total as denominator

B. The row total as denominator

C. The column total as denominator

D. The largest cell as denominator

Correct Answer: B. The row total as denominator

Question 8

A line chart is especially appropriate when the horizontal axis represents:

A. Unordered departments

B. Chronological time

C. Employee names

D. Nominal job titles

Correct Answer: B. Chronological time

Question 9

Which statement about a standard deviation is correct?

A. It is always equal to the mean

B. It is measured in squared units

C. It is the square root of variance

D. It can be calculated only for categorical data

Correct Answer: C. It is the square root of variance

Question 10

A turnover rate rises from 4% to 7%. What is the increase in percentage points?

A. 3 percentage points

B. 3% of the original rate

C. 11 percentage points

D. 75 percentage points

Correct Answer: A. 3 percentage points