Descriptive Statistics — Lecture Notes Botho University

Descriptive Statistics
Lecture Notes — Quantitative Methods
Botho University — Prepared for Student Use

Contents

1 Introduction to Descriptive Statistics 3

1.1 Two Branches of Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Populations, Samples, and Parameters . . . . . . . . . . . . . . . . . . 3
1.3 Types of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Levels of Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
End-of-Chapter Exercises — Section 1 . . . . . . . . . . . . . . . . . . . . . 5

2 Frequency Distributions 6
2.1 Constructing a Frequency Table . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Class Limits, Class Boundaries, and Midpoints . . . . . . . . . . 6
2.2 Relative and Cumulative Frequency . . . . . . . . . . . . . . . . . . . . 7
End-of-Chapter Exercises — Section 2 . . . . . . . . . . . . . . . . . . . . . 8

3 Measures of Central Tendency 9

3.1 The Arithmetic Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1 Weighted Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 The Median . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.1 Median from Grouped Data (Interpolation) . . . . . . . . . . . 11
3.3 The Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3.1 Mode from Grouped Data . . . . . . . . . . . . . . . . . . . . . 11
3.3.2 Mean from Grouped Data (Midpoint Method) . . . . . . . . . . 12
3.4 Comparing Mean, Median, and Mode . . . . . . . . . . . . . . . . . . . 12
End-of-Chapter Exercises — Section 3 . . . . . . . . . . . . . . . . . . . . . 14

4 Measures of Dispersion (Spread) 15

4.1 Range and Interquartile Range . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Quartiles and Percentiles . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3 Standard Deviation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3.1 The Empirical Rule (for Normal Distributions) . . . . . . . . . . 18
4.3.2 Z-Score (Standardisation) . . . . . . . . . . . . . . . . . . . . . 18
4.4 Coefficient of Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.5 Effect of Transformations on Measures of Spread . . . . . . . . . . . . . 19
End-of-Chapter Exercises — Section 4 . . . . . . . . . . . . . . . . . . . . . 21

5 Data Visualisation 23
5.1 Histograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2 Box Plots (Box-and-Whisker) . . . . . . . . . . . . . . . . . . . . . . . 24

1
Descriptive Statistics — Lecture Notes Botho University

5.3 Stem-and-Leaf Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

5.4 Ogives (Cumulative Frequency Curves) . . . . . . . . . . . . . . . . . . 26
5.5 Bar Charts and Pie Charts . . . . . . . . . . . . . . . . . . . . . . . . . 27
End-of-Chapter Exercises — Section 5 . . . . . . . . . . . . . . . . . . . . . 28

6 Shape of Distributions 30
6.1 Skewness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.2 Kurtosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
End-of-Chapter Exercises — Section 6 . . . . . . . . . . . . . . . . . . . . . 32

7 Exercises with Full Solutions 33

7.1 Exercises — Central Tendency and Frequency . . . . . . . . . . . . . . 33
7.2 Exercises — Dispersion, Shape, and Visualisation . . . . . . . . . . . . 34

8 Chapter Review: Mixed Exercises 35

9 Worked Solutions to Selected End-of-Chapter Exercises 38

Solutions — Section 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Solutions — Section 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Solutions — Section 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Solutions — Section 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Solutions — Section 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Solutions — Section 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

10 Practice Examination 45

11 Key Formulas Summary 48

Learning Outcomes 48

2
Descriptive Statistics — Lecture Notes Botho University

1 Introduction to Descriptive Statistics

Definition. Descriptive statistics consists of methods for organising, sum-

marising, and presenting data in a meaningful way. It describes the main features
of a dataset without making inferences or predictions beyond the data itself.

1.1 Two Branches of Statistics

• Descriptive statistics — summarises what is in the data: measures of centre
(mean, median, mode), measures of spread (variance, standard deviation), and
graphical displays (histograms, box plots).
• Inferential statistics — uses sample data to draw conclusions about a larger
population: hypothesis tests, confidence intervals, regression.

1.2 Populations, Samples, and Parameters

A population is the complete set of all individuals, items, or observations of
interest. A sample is a subset of the population selected for study. A parameter
is a numerical measure that describes a population (e.g. population mean µ); a
statistic is a numerical measure computed from a sample (e.g. sample mean x̄).

In practice, populations are often too large or inaccessible to measure in full. We

therefore collect a representative sample and use descriptive statistics to summarise
it, and inferential statistics to generalise back to the population.

Concept Symbol (Population) Symbol (Sample) Example

Mean µ x̄ Average exam score

Variance σ2 s2 Spread of salaries
Std Dev σ s Spread of heights
Size N n Number of observations

Why does notation matter?

P Using µ vs x̄ is not just cosmetic. They are com-
putedPdifferently: µ = xi /N divides by the whole population size N , whereas
x̄ = xi /n uses the sample size n. Similarly, the sample variance divides by
n − 1 (not n) to correct for bias — see Section 4.

Example 1.1 — Descriptive vs Inferential.

A hospital records the ages of 200 patients admitted in January.
• Descriptive: The mean age is 47.3 years; the most common age group is
40–49; 30% of patients are over 60.

3
Descriptive Statistics — Lecture Notes Botho University

• Inferential: Based on this sample, we estimate the mean age of all patients
admitted to this hospital is between 45.1 and 49.5 years (95% confidence
interval).

1.3 Types of Data

Qualitative (Categorical) data describes categories or labels with no natural
numeric meaning.
Quantitative (Numerical) data takes numeric values for which arithmetic op-
erations make sense.

Quantitative data is further divided:

• Discrete: countable values with gaps — number of cars sold (3, 7, 12), number of
students in a class.
• Continuous: any value in a range — height (1.742 m), weight (63.4 kg), time
(14.72 s).

Example 1.2 — Classifying Variables.

Variable Type Reason

Exam score (0–100) Quantitative, continuous Any value in range

Number of siblings Quantitative, discrete Countable, integer only
Favourite colour Qualitative (categorical) Labels, not numbers
Student ID number Qualitative (nominal) Numbers used as labels
Pain rating (1–10) Qualitative (ordinal) Ordered, but gaps unequal

1.4 Levels of Measurement

The four levels, in order of increasing mathematical power:

Level Properties Valid Averages Examples

Nominal Categories only Mode only Blood type, gender,

country
Ordinal Order meaningful Mode, median Satisfaction scale,
exam grades
Interval Equal gaps, no true zero Mean, median, mode Temperature (°C), IQ,
year
Ratio Equal gaps, true zero All Height, income, age,
speed

4
Descriptive Statistics — Lecture Notes Botho University

Example 1.3 — Interval vs Ratio.

Room temperature: 20°C vs 40°C.
• The difference of 20°C is meaningful (interval property).
• We cannot say 40°C is “twice as hot” as 20°C — there is no true zero on
the Celsius scale.
• At the Ratio scale: BWP 20,000 is twice BWP 10,000 because zero income
(BWP 0) is meaningful.

Why does the level of measurement matter? The level of measurement

determines which statistical operations and summaries are valid. For example:
• Computing the mean of jersey numbers (nominal) is meaningless — the
average of jersey 7 and jersey 11 is not “jersey 9”.
• Saying one temperature is “twice” another (interval) is invalid without a
true zero.
• Always identify the measurement level before choosing a summary statistic
or chart type.

End-of-Chapter Exercises — Section 1

Exercise 1.1. Classify each of the following variables as qualitative or quantita-
tive, and if quantitative, state whether it is discrete or continuous.
(a) The number of goals scored in a football match.
(b) The time (in minutes) taken to complete an assignment.
(c) The colour of a car in a university car park.
(d) The monthly salary (BWP) of an employee.
(e) The number of ATM transactions processed per day.
(f) A customer satisfaction rating: Poor / Fair / Good / Excellent.

Exercise 1.2. For each variable below, identify the level of measurement (Nom-
inal, Ordinal, Interval, or Ratio) and justify your choice.
(a) The year a student was born.
(b) A patient’s blood pressure (mmHg).
(c) Military rank (Private, Corporal, Sergeant, . . . ).
(d) Country of birth.
(e) Daily maximum temperature recorded in Gaborone (°C).
(f) Distance driven (km) during a delivery route.

5
Descriptive Statistics — Lecture Notes Botho University

Exercise 1.3. A researcher surveys 500 households in Francistown and records:

household size, monthly income (BWP), type of dwelling (house/flat/room), and
the head of household’s highest education level.
(a) Classify each of the four variables by type and level of measurement.
(b) Which variables can be meaningfully added or averaged? Explain.
(c) For the education level variable, which measure of central tendency is most
appropriate and why?

Exercise 1.4 (Discussion). A company reports that the “average” annual bonus
paid to employees was BWP 12,000. The CEO received a bonus of BWP 500,000.
(a) Which branch of statistics — descriptive or inferential — does this statement
involve?
(b) Explain why the mean may be a misleading summary here.
(c) What alternative measure would better represent the “typical” employee
bonus?

2 Frequency Distributions

Definition. A frequency distribution is a table that groups data values into

classes (bins) and records the number of observations (frequency) in each class.

2.1 Constructing a Frequency Table

Steps:
1. Find the range: R = Max − Min.
√
2. Choose the number of classes k. Rule of thumb: k ≈ n (use 5–15 classes).
3. Calculate class width: w = R/k (round up to a convenient value).
4. Set class boundaries and tally observations into each class.

2.1.1 Class Limits, Class Boundaries, and Midpoints

Class limits are the stated endpoints of a class as they appear in the table (e.g.
40–49).
Class boundaries are the true continuous boundaries used for computation,
found by averaging adjacent limits: e.g. for class 40–49 followed by 50–59, the
upper boundary is (49 + 50)/2 = 49.5.
Class midpoint (or class mark) is the average of the lower and upper class limits:
midpoint = (L + U )/2.

6
Descriptive Statistics — Lecture Notes Botho University

Class boundaries eliminate gaps between classes for continuous data. The midpoint
is used to represent all values within a class when computing the grouped mean and
other statistics.
Common mistakes when constructing frequency tables:
• Overlapping classes — e.g. 40–50 and 50–60 leave ambiguity about where
50 belongs. Use 40–49 and 50–59 for discrete data, or use strict inequalities
(40 ≤ x < 50) for continuous data.
• Unequal class widths — unless justified (e.g. open-ended classes), all widths
should be equal to allow fair visual comparison in histograms.
• Too few or too many classes — fewer than 5 classes lose detail; more than
15 make the table hard to interpret.
• Not tallying carefully — always recount to ensure
P
fi = n.

2.2 Relative and Cumulative Frequency

fi
Relative frequency: rfi = (1)
n
j
X
Cumulative frequency: cfj = fi (2)
i=1
j
X
Cumulative relative frequency: crfj = rfi (3)
i=1

Example 2.1 — Exam Scores (n = 20).

Raw data (sorted): 45, 52, 58, 61, 63, 66, 68, 70, 71, 72, 74, 75, 76, 78, 80, 82, 84,
87, 91, 95
√
R = 95 − 45 = 50, k = ⌈ 20⌉ = 5, w = 50/5 = 10

Class Midpoint Freq f Rel Freq rf Cum Freq cf Cum Rel Freq

45 – 54 49.5 2 0.10 2 0.10

55 – 64 59.5 3 0.15 5 0.25
65 – 74 69.5 6 0.30 11 0.55
75 – 84 79.5 6 0.30 17 0.85
85 – 94 89.5 2 0.10 19 0.95
95 – 104 99.5 1 0.05 20 1.00

Total 20 1.00

Interpretations:

7
Descriptive Statistics — Lecture Notes Botho University

• 85% of students scored below 85 (cf = 17 at the end of the 75–84 class).
• The distribution is roughly symmetric, peaking in the 65–74 and 75–84
classes.
• Only 5% of students scored 95 or above.

Example 2.2 — Constructing a Frequency Table from Raw Data (Monthly

Expenditure).
The following are the monthly personal expenditures (BWP) for 25 students:
560, 620, 480, 710, 540, 595, 670, 720, 500, 630,
610, 490, 740, 580, 655, 515, 690, 560, 600, 735,
470, 625, 680, 545, 615
Step 1: Min = 470, Max = 740. Range R = 740 − 470 = 270.
√
Step 2: k ≈ 25 = 5 classes.
Step 3: Width w = 270/5 = 54 → round up to 60 for convenience.
Step 4: Start first class at 450 (below the minimum).

Class (BWP) Midpoint Freq f Rel Freq Cum Freq Cum Rel Freq

450 – 509 479.5 4 0.16 4 0.16

510 – 569 539.5 6 0.24 10 0.40
570 – 629 599.5 8 0.32 18 0.72
630 – 689 659.5 5 0.20 23 0.92
690 – 749 719.5 2 0.08 25 1.00

Total 25 1.00

Interpretations:
• 40% of students spend less than BWP 570 per month.
• The modal class is 570–629 (highest frequency of 8).
• The distribution appears slightly right-skewed, with fewer students in the
high-expenditure classes.

End-of-Chapter Exercises — Section 2

Exercise 2.1. The ages (years) of 18 employees at a company are:
23, 27, 31, 45, 29, 38, 52, 41, 34,
26, 30, 48, 37, 55, 23, 44, 31, 39
(a) Determine a suitable class width and number of classes.

8
Descriptive Statistics — Lecture Notes Botho University

(b) Construct a complete frequency distribution table with relative and cumu-
lative frequencies.
(c) What percentage of employees are younger than 40?
(d) Identify the modal class.

Exercise 2.2. The table below shows the distribution of monthly data usage
(GB) for 50 mobile subscribers.

Class (GB) Frequency Cumulative Frequency

0 – 4 8 8
5 – 9 14 22
10 – 14 16 38
15 – 19 9 47
20 – 24 3 50

(a) Complete the relative frequency and cumulative relative frequency columns.
(b) What proportion of subscribers use between 5 and 14 GB?
(c) What is the minimum data usage above which only 12% of subscribers fall?
(d) Sketch a rough histogram for this distribution and describe its shape.

Exercise 2.3. A lecturer records the following scores out of 50 for a class test (n
= 30):
18, 22, 35, 41, 29, 38, 45, 27, 33, 48,
20, 36, 42, 30, 25, 39, 47, 28, 34, 40,
23, 37, 43, 31, 26, 32, 46, 21, 44, 38
(a) Using 6 classes of equal width, construct the frequency distribution.
(b) Calculate the relative frequency for each class.
(c) What percentage of students scored 35 or above?
(d) Compare your modal class with the actual mode. Do they agree? Explain
any discrepancy.

3 Measures of Central Tendency

Measures of central tendency locate the centre of a dataset.

3.1 The Arithmetic Mean

9
Descriptive Statistics — Lecture Notes Botho University

PN Pn
i=1 xi i=1 xi
Population mean: µ = Sample mean: x̄ = (4)
N n
Pn
Property 3.1. The sum of deviations from the mean is always zero: i=1 (xi − x̄) = 0.
Weaknesses: The mean is sensitive to outliers (extreme values). A single large value
can substantially raise or lower x̄.

3.1.1 Weighted Mean

When observations carry different levels of importance (weights wi ):
Pn
w i xi
x̄w = Pi=1
n (5)
i=1 wi

Example 3.1 — Arithmetic Mean (Daily Sales).

Daily BWP sales: 120, 145, 98, 200, 155, 145, 175, 130, 145, 188

120 + 145 + 98 + 200 + 155 + 145 + 175 + 130 + 145 + 188 1501
x̄ = = = 150.1 BWP
10 10

Example 3.2 — Weighted Mean (GPA).

Course Credits (wi ) Grade (xi ) wi xi

Maths 4 3.7 14.8
Statistics 3 3.3 9.9
English 2 3.0 6.0
Total 9 30.7

14.8 + 9.9 + 6.0 30.7

x̄w = = ≈ 3.41
4+3+2 9
The GPA is 3.41, closer to the Maths grade because it carries the most credit
weight.

3.2 The Median

Definition. The median is the middle value of an ordered dataset.
n+1
• Odd n: median = value at position
2
n n
• Even n: median = average of values at positions and + 1
2 2

The median is robust — not affected by outliers. Preferred for skewed distributions
(income, house prices).

10
Descriptive Statistics — Lecture Notes Botho University

3.2.1 Median from Grouped Data (Interpolation)

n
− cf
M =L+ 2 ×w (6)
f
where L = lower boundary of median class, cf = cumulative frequency before the
median class, f = frequency of median class, w = class width.

Example 3.3 — Median from Grouped Data.

Using the exam score table (n = 20), n/2 = 10.
The cumulative frequency first reaches or exceeds 10 in the class 65–74 (cf before
= 5, f = 6, w = 10, L = 65):
10 − 5 50
M = 65 + × 10 = 65 + = 65 + 8.33 = 73.33
6 6
Interpretation: Half the students scored below approximately 73.3 on the exam.

3.3 The Mode

Definition. The mode is the value (or class) that occurs most frequently. A
dataset may be unimodal, bimodal, multimodal, or have no mode.

Example 3.4 — Mode and Bimodal Distribution.

(a) Unimodal: Sales data 120, 145, 98, 200, 155, 145, 175, 130, 145, 188, 145
⇒ Mode = 145 (appears 3 times)
(b) Bimodal: Test scores 55, 55, 60, 72, 75, 80, 80, 85 ⇒ Modes = 55 and 80
A bimodal distribution often indicates two distinct subgroups (e.g. high-achievers
and low-achievers).

3.3.1 Mode from Grouped Data

When data is presented in a frequency table, the exact mode is unknown. We can
estimate it using the modal class (the class with the highest frequency) and the
following interpolation formula:

d1
Mode ≈ L + ×w (7)
d1 + d 2
where L = lower boundary of the modal class, d1 = difference between the modal
class frequency and the frequency of the class before it, d2 = difference between
the modal class frequency and the frequency of the class after it, and w = class
width.

Example 3.4b — Mode from Grouped Data.

Using the exam score table from Example 2.1 (n = 20):

11
Descriptive Statistics — Lecture Notes Botho University

Class Midpoint Freq f

45–54 49.5 2
55–64 59.5 3
65–74 69.5 6 ← modal class
75–84 79.5 6
85–94 89.5 2
95–104 99.5 1

The modal class is 65–74 (tied with 75–84, but 65–74 comes first). Using L = 65,
d1 = 6 − 3 = 3, d2 = 6 − 6 = 0, w = 10:
3
Mode ≈ 65 + × 10 = 65 + 10 = 75
3+0

Note: When two classes tie for highest frequency (as here), the convention is
to select the first modal class. The result suggests the estimated mode lies at
approximately 75 — at the upper end of the first modal class, being “pulled”
toward the equally frequent adjacent class.

3.3.2 Mean from Grouped Data (Midpoint Method)

When raw data is unavailable and only a frequency table is given, the mean is estimated
using class midpoints:
P P
fi · m i fi · m i
x̄ ≈ P = (8)
fi n
where mi is the midpoint of class i and fi is its frequency.

Example 3.4c — Mean from Grouped Data.

Using the exam score table (Example 2.1):

2(49.5) + 3(59.5) + 6(69.5) + 6(79.5) + 2(89.5) + 1(99.5) 99 + 178.5 + 417 + 477 + 179 + 99
x̄ ≈ =
20 20

The estimated mean is 72.5. This is an approximation — the actual mean (from
the raw data) is 72.9 — showing the midpoint method introduces only a small
error.

3.4 Comparing Mean, Median, and Mode

12
Descriptive Statistics — Lecture Notes Botho University

Measure Best Used When Limitations

Mean Symmetric data, no outliers, interval/ratio scale Distorted by extreme values

Median Skewed data, outliers present, ordinal or ratio Does not use all values
Mode Nominal data, bimodal identification May not be unique; not always
representative

Relationship in Skewed Distributions:

• Right (positive) skew: Mode < Median < Mean (tail pulls mean right)
• Left (negative) skew: Mean < Median < Mode (tail pulls mean left)
• Symmetric: Mean ≈ Median ≈ Mode

Example 3.5 — Effect of an Outlier (House Prices, BWP thousands).

Dataset: 180, 195, 200, 210, 220, 225, 230, 850

180 + 195 + 200 + 210 + 220 + 225 + 230 + 850 2310

x̄ = = = 288.75
8 8
220 + 225
Median = = 222.5
2

The mansion (BWP 850k) pulls the mean up by BWP 66k above the median.
The median is the better representation of the “typical” house price. This is a
right-skewed distribution: Mean > Median.

Example 3.6 — Complete Central Tendency Analysis (Employee Salaries,

BWP thousands).
A company’s HR department records the following monthly salaries (BWP ’000)
for 12 employees:
8, 9, 9, 10, 11, 12, 12, 12, 14, 15, 18, 42
Mean:
8 + 9 + 9 + 10 + 11 + 12 + 12 + 12 + 14 + 15 + 18 + 42 172
x̄ = = ≈ 14.33 BWP ’000
12 12

Median (even n = 12, average of 6th and 7th values):

12 + 12
Median = = 12 BWP ’000
2

Mode: 12 BWP ’000 (appears 3 times)

Analysis:

13
Descriptive Statistics — Lecture Notes Botho University

• Mean (14.33) > Median (12) > Mode (12): right-skewed distribution.
• The outlier salary of BWP 42,000 inflates the mean by over BWP 2,000.
• The median (BWP 12,000) most accurately represents the typical em-
ployee’s salary.
• This pattern is common in wage data: a few high earners pull the mean far
above what most workers earn.

End-of-Chapter Exercises — Section 3

Exercise 3.1. A shop records the following number of customers served per hour
over 12 hours:
14, 18, 22, 9, 31, 18, 25, 18, 12, 27, 20, 15
(a) Calculate the mean, median, and mode.
(b) The shop manager replaces the value 31 with the correct value 13 (data
entry error). Recalculate all three measures.
(c) Which measure changed the most? What does this tell you about the sen-
sitivity of each measure to a single extreme value?

Exercise 3.2. The grouped frequency table below shows the weekly study hours
of 40 university students.

Hours Frequency Midpoint

0 – 4 3 2
5 – 9 8 7
10 – 14 14 12
15 – 19 10 17
20 – 24 5 22

(a) Use the midpoint method to estimate the mean study hours.
(b) Use interpolation to find the median study hours.
(c) Identify the modal class.
(d) The university recommends a minimum of 12 study hours per week. Esti-
mate the percentage of students meeting this minimum.

Exercise 3.3 (Weighted Mean). A student takes four modules in a semester.

Their results are shown below.

14
Descriptive Statistics — Lecture Notes Botho University

Module Credits Score (%)

Business Mathematics 5 72
Communication Skills 2 85
Introduction to IT 3 68
Accounting Fundamentals 4 78

(a) Calculate the weighted mean score (weighted by credits).

(b) Calculate the simple (unweighted) mean.
(c) Which is higher? Explain why, based on the data.
(d) If the student wants a weighted average of at least 75%, what minimum
score must they achieve in a new 3-credit module?

Exercise 3.4. The mean salary of 8 analysts at a firm is BWP 18,500 per month.
A new analyst joins with a salary of BWP 22,000.
(a) What is the new mean salary of all 9 analysts?
(b) If instead the new hire’s salary was BWP 14,000, what would the new mean
be?
(c) A department has 5 male employees with a mean salary of BWP 16,000
and 3 female employees with a mean salary of BWP 19,500. What is the
combined mean salary for the department?

Exercise 3.5 (Challenge). For a dataset of n values, state whether each claim
is always true, sometimes true, or never true, and justify with an example.
(a) Removing the largest value will always decrease the mean.
(b) The median can never be greater than the mean.
(c) A dataset can have more than one mode.
(d) If x̄ = Median, the distribution is perfectly symmetric.

4 Measures of Dispersion (Spread)

Measures of central tendency alone are insufficient — two datasets can share the same
mean but differ vastly in spread.

4.1 Range and Interquartile Range

15
Descriptive Statistics — Lecture Notes Botho University

Range = Max − Min (9)

IQR = Q3 − Q1 (10)
Outlier fences : Q1 − 1.5 × IQR and Q3 + 1.5 × IQR (11)

Example 4.1 — Range and IQR (Maths Test, n = 10).

Ordered data: 10, 11, 12, 13, 14, 15, 16, 17, 18, 18
• Range = 18 − 10 = 8
• Q1 = median of lower half {10, 11, 12, 13, 14} = 12
• Q3 = median of upper half {15, 16, 17, 18, 18} = 17
• IQR = 17 − 12 = 5
• Lower fence = 12 − 1.5(5) = 4.5 Upper fence = 17 + 1.5(5) = 24.5
• All values lie within [4.5, 24.5] ⇒ no outliers
Five-number summary: Min=10, Q1 =12, Median=14.5, Q3 =17, Max=18

4.2 Quartiles and Percentiles

Quartiles divide ordered data into four equal parts. Percentiles divide ordered
data into 100 equal parts. The p-th percentile Pp is the value below which p% of
the data falls.
Relationship: Q1 = P25 , Q2 = Median = P50 , Q3 = P75 .

Locating the p-th percentile in raw data (n observations):

1. Sort the data in ascending order.
p
2. Compute the locator : L = × n.
100
3. If L is not a whole number, round up to the next integer. Pp is that value.
4. If L is a whole number, Pp is the average of the L-th and (L + 1)-th values.

Example 4.1b — Quartiles and Percentiles.

Data (n = 12, sorted): 5, 8, 10, 12, 14, 15, 17, 19, 21, 24, 28, 33
Q1 (P25 ): L = (25/100) × 12 = 3. Since L is a whole number, Q1 = (x3 + x4 )/2 =
(10 + 12)/2 = 11.
Q2 (Median, P50 ): L = (50/100)×12 = 6. Q2 = (x6 +x7 )/2 = (15+17)/2 = 16.
Q3 (P75 ): L = (75/100) × 12 = 9. Q3 = (x9 + x10 )/2 = (21 + 24)/2 = 22.5.
P80 : L = (80/100) × 12 = 9.6. Round up to 10. P80 = x10 = 24.
P90 : L = (90/100) × 12 = 10.8. Round up to 11. P90 = x11 = 28.

16
Descriptive Statistics — Lecture Notes Botho University

Interpretation: 80% of values in this dataset fall below 24, and 90% fall below
28.
IQR = Q3 − Q1 = 22.5 − 11 = 11.5.
Lower fence = 11 − 1.5(11.5) = −6.25; upper fence = 22.5 + 1.5(11.5) = 39.75.
All values lie within [−6.25, 39.75] ⇒ no outliers.

Caution — Different Quartile Methods: Different textbooks and software

packages use slightly different rules for locating quartiles. Excel, for example, uses
a different formula from the one above. The differences are small and diminish as
n grows, but you should always state which method you are using when precision
matters. In this course, use the locator method above.

PN
2 i=1 (xi − µ)2
Population variance: σ = (12)
N
Pn
2 i=1 (xi− x̄)2
Sample variance: s = (13)
n−1
xi )2
P
x2i − (
P
2 n
Computational shortcut: s = (14)
n−1

Why divide by n − 1? Using n − 1 (Bessel’s correction) produces an unbiased

estimator of the population variance. A sample tends to underestimate the true
spread because it is constrained to its own mean; dividing by n − 1 corrects this.

Example 4.2 — Sample Variance (data: 2, 4, 4, 4, 5, 5, 7, 9).

2+4+4+4+5+5+7+9 40
x̄ = = =5
8 8

xi xi − x̄ (xi − x̄)2
2 −3 9
4 −1 1
4 −1 1
4 −1 1
5 0 0
5 0 0
7 +2 4
9 +4 16
Total 0 32

17
Descriptive Statistics — Lecture Notes Botho University

32 32 32
s2 = = ≈ 4.57 σ2 = = 4.00
8−1 7 8
P
Note: (xi − x̄) = 0 confirms the calculation is correct.

4.3 Standard Deviation

√
rP
(xi − x̄)2
s = s2 = (15)
n−1

The standard deviation has the same units as the original data, making it directly
interpretable.

4.3.1 The Empirical Rule (for Normal Distributions)

For a roughly bell-shaped distribution:
• x̄ ± 1s contains approximately 68% of observations
• x̄ ± 2s contains approximately 95% of observations
• x̄ ± 3s contains approximately 99.7% of observations

4.3.2 Z-Score (Standardisation)

x − x̄
z= (16)
s

The z-score measures how many standard deviations x is from the mean. z > 0: above
mean; z < 0: below mean.

Example 4.3 — Standard Deviation and Z-Score.

(a) Standard deviation from Example 4.2:
√
s = 4.57 ≈ 2.14

68% of values lie in [5 − 2.14, 5 + 2.14] = [2.86, 7.14].

(b) Z-score: A student scores 82 on a test where x̄ = 70, s = 8.
82 − 70 12
z= = = 1.5
8 8
The student is 1.5 standard deviations above average — in approximately the top
7% of the class.
(c) Below average: A student scores 58.
58 − 70 −12
z= = = −1.5
8 8
Symmetrically, this student is 1.5 SDs below the mean.

18
Descriptive Statistics — Lecture Notes Botho University

4.4 Coefficient of Variation

s
CV = × 100% (17)
x̄

The CV is dimensionless — it expresses variability as a percentage of the mean,

allowing comparison across datasets with different units or scales.

Example 4.4 — Investment Risk Comparison.

Fund Mean Return Std Dev (s) CV Verdict

Fund A 12% 3% 25% More consistent
Fund B 20% 7% 35% Higher absolute return, more relative risk
Fund B earns more on average, but for every unit of return earned, you accept
more relative risk. Fund A is the safer relative choice.
Cross-unit comparison: Heights (x̄ = 170 cm, s = 8 cm) vs weights (x̄ = 70
kg, s = 12 kg).
8 12
CVheight = × 100 = 4.7% CVweight = × 100 = 17.1%
170 70
Weight is relatively more variable than height in this group.

4.5 Effect of Transformations on Measures of Spread

A common question in statistics is: what happens to the mean and standard deviation
if we add a constant to every value, or multiply every value by a constant?

Let yi = a + b xi (a linear transformation of the data). Then:

ȳ = a + b x̄ (18)
sy = |b| · sx (19)
s2y = b2 · s2x (20)

Key rules to remember:

• Adding or subtracting a constant (b = 1): shifts the mean by that
constant but does not change the standard deviation or variance. The spread
is unchanged.
• Multiplying by a constant (a = 0): scales the mean and the standard
deviation by |b|; the variance scales by b2 .
• The CV changes when you multiply (since both mean and SD scale equally,
the ratio is preserved), but changes when you add a constant (the SD stays
the same but the mean shifts).

19
Descriptive Statistics — Lecture Notes Botho University

Example 4.6 — Transformations (Converting Temperature).

A dataset of temperatures in Celsius has x̄C = 25°C and sC = 4°C.
Convert to Fahrenheit using F = 32 + 1.8 C:

F̄ = 32 + 1.8(25) = 32 + 45 = 77°F
sF = 1.8 × 4 = 7.2°F

Practical test: If a teacher adds 5 bonus marks to every student’s score (a = 5,

b = 1):
• New mean = x̄ + 5 (shifts up)
• New standard deviation = s (unchanged — all scores shift equally, so spread
is preserved)
• New variance = s2 (unchanged)
If instead scores are scaled by multiplying by 1.1 (a = 0, b = 1.1):
• New mean = 1.1x̄
• New standard deviation = 1.1s
• New variance = 1.21s2

Example 4.5 — Full Dispersion Analysis (Monthly Rainfall, mm).

Gaborone’s monthly rainfall (mm) over 10 months:
12, 0, 5, 85, 120, 95, 60, 8, 3, 42
Step 1 — Order the data: 0, 3, 5, 8, 12, 42, 60, 85, 95, 120
Step 2 — Five-number summary:
• Min = 0, Max = 120
• Q1 = median of {0, 3, 5, 8, 12} = 5
• Q3 = median of {42, 60, 85, 95, 120} = 85
• Median = (12 + 42)/2 = 27
• IQR = 85 − 5 = 80
• Fences: 5 − 120 = −115 (lower) and 85 + 120 = 205 (upper). No outliers.
Step 3 — Mean and standard deviation:
430
x̄ = = 43 mm
10

20
Descriptive Statistics — Lecture Notes Botho University

xi xi − x̄ (xi − x̄)2
0 −43 1849
3 −40 1600
5 −38 1444
8 −35 1225
12 −31 961
42 −1 1
60 +17 289
85 +42 1764
95 +52 2704
120 +77 5929
Total 0 17766

17766
s2 = ≈ 1974 mm2 s ≈ 44.4 mm
9
Step 4 — CV:
44.4
CV = × 100 ≈ 103%
43
A CV above 100% confirms that this rainfall data is extremely variable relative
to its mean — a feature common in arid climates where rainfall is concentrated
in short, intense wet seasons.

End-of-Chapter Exercises — Section 4

Exercise 4.1. The following are the waiting times (minutes) recorded at two
bank branches on a particular morning.
Branch A: 4, 6, 7, 5, 8, 6, 9, 5, 7, 8
Branch B: 2, 3, 10, 1, 12, 4, 9, 11, 2, 6
(a) Calculate the range and IQR for each branch.
(b) Calculate the sample variance and standard deviation for each branch.
(c) Calculate the coefficient of variation for each branch.
(d) Which branch offers a more consistent service? Justify your answer using
two different measures.
(e) Check each branch for outliers using the 1.5 × IQR rule.

Exercise 4.2 (Z-Scores). In an end-of-semester examination, the class mean is

63 and the standard deviation is 9.
(a) A student scored 78. Calculate and interpret her z-score.

21
Descriptive Statistics — Lecture Notes Botho University

(b) Another student scored 48. Calculate and interpret his z-score.
(c) Using the Empirical Rule, between what two scores do approximately 95%
of students fall?
(d) What score corresponds to a z-score of −2.5?
(e) Two students, Thabo and Kefilwe, take different modules. Thabo scores 72
in Finance (x̄ = 65, s = 10) and Kefilwe scores 80 in Marketing (x̄ = 75,
s = 8). Who performed better relative to their class? Show your working.

Exercise 4.3 (CV and Comparison).

Product Mean Daily Sales (units) Std Dev CV

Product X 200 40 ?
Product Y 50 15 ?
Product Z 500 120 ?

(a) Calculate the CV for each product.

(b) Rank the products from least to most variable relative to their mean.
(c) A manager argues that Product Z is the most risky because it has the highest
standard deviation. Do you agree? Explain.
(d) If Product X’s mean drops to 80 units while its standard deviation stays
at 40, recalculate its CV. What does this tell you about the relationship
between mean and relative variability?

Exercise 4.4 (Computational Shortcut). Use the computational shortcut

formula P 2 (P xi )2
xi − n
s2 =
n−1
to find the sample variance and standard deviation for the dataset:

3, 7, 7, 9, 10, 12, 15
(xi −x̄)2
P
Verify your answer using the definition formula s2 = n−1
.

Exercise 4.5 (Discussion). Explain, in your own words:

(a) Why the standard deviation is more useful than the variance when describing
data spread to a non-technical audience.
(b) Why dividing by n − 1 rather than n gives an unbiased estimate of the
population variance.
(c) In what situation would you prefer the IQR over the standard deviation as
a measure of spread?

22
Descriptive Statistics — Lecture Notes Botho University

Exercise 4.6 (Transformations).

A dataset of 20 employees’ monthly salaries has x̄ = BWP 14,200 and s =
BWP 3,100.
(a) The company gives every employee a flat raise of BWP 800 per month.
What are the new mean and standard deviation?
(b) Instead, the company gives a 10% raise (multiply every salary by 1.10).
What are the new mean and standard deviation?
(c) Under which raise is the spread (in BWP) larger? Explain intuitively why
this makes sense.
(d) A statistician converts all salaries from BWP to US dollars using the ex-
change rate 1 USD = 13.5 BWP. What are the mean and standard deviation
in USD? What is the CV, and how does it compare to the original CV in
BWP?

Exercise 4.7 (Quartiles and Percentiles).

The following are the monthly electricity bills (BWP) for 16 households, sorted
in ascending order:
85, 102, 115, 128, 134, 141, 155, 163, 170, 182, 195, 208, 224, 237, 260, 315
(a) Find Q1 , Q2 (median), and Q3 using the locator method.
(b) Calculate the IQR and check for outliers.
(c) Find P30 , P70 , and P90 .
(d) Interpret P90 in the context of this data.
(e) Between which two values do the middle 50% of bills fall? What is this
range called?

5 Data Visualisation

5.1 Histograms
A histogram displays the frequency (or relative frequency) of continuous data within
class intervals. Key features:
• Bars are adjacent (no gaps) — reflects continuous data.
• Bar height = frequency (or relative frequency); bar width = class width.
• The shape of the histogram reveals the distribution: symmetric, right-skewed,
left-skewed, bimodal, or uniform.

23
Descriptive Statistics — Lecture Notes Botho University

Example 5.1 — Histogram for Exam Scores.

Exam Score Distribution (n=20)
8

6
Frequency
4

0
45 55 65 75 85 95
Score

The histogram shows a roughly symmetric distribution peaking in the 65–84 range,
with light tails on both sides.

5.2 Box Plots (Box-and-Whisker)

A box plot graphically displays the five-number summary: Min, Q1 , Median, Q3 ,
Max.
IQR
Median

Min Q1 Q3 Max

Example 5.2 — Box Plot (Maths Scores, n=10).

Five-number summary: Min=10, Q1 =12, Median=14.5, Q3 =17, Max=18
• IQR = 17 − 12 = 5
• Lower fence = 4.5; upper fence = 24.5
• All data within fences ⇒ no outliers
• The median (14.5) is slightly closer to Q1 (12) than Q3 (17), suggesting a
mild right skew

5.3 Stem-and-Leaf Plots

A stem-and-leaf plot (or stem plot) is a quick way to display the shape of a small-
to-medium dataset while retaining the actual data values. Each data value is split into
a stem (leading digit(s)) and a leaf (final digit).
Advantages over a histogram:
• The original data values are preserved — the plot itself is the data.

24
Descriptive Statistics — Lecture Notes Botho University

• Easy to identify the median, quartiles, and outliers by inspection.

• Fast to construct by hand for n ≤ 50.

Example 5.0a — Stem-and-Leaf Plot (Exam Scores, n = 20).

Raw data: 45, 52, 58, 61, 63, 66, 68, 70, 71, 72, 74, 75, 76, 78, 80, 82, 84, 87, 91,
95
Stem Leaves
4 5
5 28
6 1368
7 0124568
8 0247
9 15
Key: Stem = tens digit, Leaf = units digit. E.g. stem 7, leaf 4 ⇒ score 74.
Reading the plot:
• n = 20: count all leaves.
• Median: average of 10th and 11th values = (72 + 74)/2 = 73. Compare with
the interpolated value of 73.3 from Example 3.3.
• Shape: the distribution is roughly bell-shaped, peaking in the 70s.
• The single value in the 40s (45) and 90s (91, 95) are in the tails.

Example 5.0b — Back-to-Back Stem-and-Leaf Plot (Comparing Two

Groups).
A teacher compares test scores for two classes (n = 10 each):
Class A: 55, 58, 62, 64, 68, 71, 74, 77, 82, 85
Class B: 48, 53, 56, 61, 63, 65, 70, 73, 79, 88
Class A (leaves, read right to left) Stem Class B (leaves)
4 8
8 5 5 36
84 2 6 135
74 1 7 039
5 2 8 8
Reading a back-to-back plot: Class A leaves are read from the stem outward
to the left; Class B leaves are read from the stem outward to the right.
Comparison:
• Class A median = (68 + 71)/2 = 69.5; Class B median = (63 + 65)/2 = 64.
• Class A generally scores higher and its distribution is more spread through
the 60s–80s.
• Class B has a low outlier at 48 and a high score of 88, suggesting more

25
Descriptive Statistics — Lecture Notes Botho University

variability in the extremes.

5.4 Ogives (Cumulative Frequency Curves)

An ogive (pronounced “oh-jive”) is a graph of the cumulative frequency (or cu-
mulative relative frequency) plotted against the upper class boundary. It allows us to
read off percentiles and medians directly from the graph.
How to draw an ogive:
1. Plot a point at the lower boundary of the first class with cumulative frequency =
0 (the ogive starts at zero).
2. For each class, plot a point at the upper class boundary with the corresponding
cumulative frequency.
3. Connect the points with straight lines.
Reading from an ogive:
• Median: find the value on the horizontal axis corresponding to cumulative fre-
quency = n/2 (or 50% on the relative ogive).
• Quartiles: Q1 at 25%, Q3 at 75%.
• Percentiles: the p-th percentile is the value at which p% of the data falls below.

Example 5.3b — Ogive for Exam Scores.

Using the cumulative frequencies from Example 2.1 (n = 20):

Upper Boundary Cumulative Frequency

44.5 0
54.5 2
64.5 5
74.5 11
84.5 17
94.5 19
104.5 20

26
Descriptive Statistics — Lecture Notes Botho University

Ogive — Exam Scores (n=20)

20

Cumulative Frequency
19
17
15

11 n/2 = 10
10

5
2
0
≈ 74.5
.5

.5

.5

.5

.5

.5

5
4.
44

54

64

74

84

94

10
Score (upper boundary)

Reading the graph: the cumulative frequency of 10 (i.e. n/2) corresponds to a score
of approximately 74.5, confirming the interpolated median of 73.3 from Example
3.3.

5.5 Bar Charts and Pie Charts

Bar charts display categorical data. Unlike histograms, bars have gaps.
• Clustered: compare multiple groups side by side.
• Stacked: show part-to-whole composition within groups.
Pie charts show proportions of a whole. Each slice angle:

fi
θi = × 360 (21)
n

Best used for 3–6 categories only. Always label both name and percentage.

Example 5.3 — Student Transport Mode (n=60).

Mode Freq Rel Freq Angle Bar height (%)

Walk 20 33% 120° 33
Bus 25 42% 151° 42
Car 12 20% 72° 20
Bicycle 3 5% 18° 5
Total 60 100% 361°* 100
*rounding error; bus slice = (25/60) × 360 = 150 exactly
Bus is the most common transport, used by 42% of students. A bar chart conveys
this clearly; a pie chart also works well since there are only 4 categories.

27
Descriptive Statistics — Lecture Notes Botho University

Example 5.4 — Interpreting and Comparing Box Plots.

Two delivery companies report the following delivery times (minutes) for 10 orders
each:
Company A (ordered): 18, 22, 25, 27, 30, 32, 35, 38, 42, 50
Company B (ordered): 10, 12, 28, 29, 30, 31, 32, 35, 36, 60
Company A: Q1 = 25, Median=31, Q3 = 38, IQR=13.
Fences: [5.5, 57.5]. All values within range — no outliers.
Company B: Q1 = 12, Median=30.5, Q3 = 35, IQR=23.
Fences: [−22.5, 69.5]. All within range — no outliers, though 10 and 60 are close
to boundaries.
Comparison:
• Both companies have similar medians (≈ 30 min), suggesting similar typical
delivery times.
• Company B has a much larger IQR (23 vs 13): its delivery times are far
more inconsistent.
• Company A’s box plot would appear more compact; Company B’s would
show a longer, wider box with longer whiskers.
• A customer who values reliability would prefer Company A; Company B’s
best case (10 min) is better but its worst case (60 min) is far worse.

End-of-Chapter Exercises — Section 5

Exercise 5.0 (Stem-and-Leaf Plots).
The following are the ages of 18 participants in a community health programme:
23, 45, 31, 52, 28, 47, 33, 61, 38, 44, 55, 29, 42, 67, 36, 51, 48, 34
(a) Construct a stem-and-leaf plot for this data.
(b) Use the plot to find the median and the quartiles.
(c) Describe the shape of the distribution based on the plot.
(d) A second group of 10 participants has ages: 19, 25, 30, 35, 38, 42, 50, 58, 63,
71. Construct a back-to-back stem-and-leaf plot comparing the two groups
and comment on the differences.

Exercise 5.1 (Histogram Interpretation). A histogram of weekly household

spending (BWP) shows the following bar heights (frequencies) for 80 households:

28
Descriptive Statistics — Lecture Notes Botho University

Class (BWP) Frequency

100 – 199 5
200 – 299 12
300 – 399 28
400 – 499 22
500 – 599 10
600 – 699 3

(a) Sketch the histogram, clearly labelling both axes.

(b) Describe the shape of the distribution (symmetric, skewed, etc.).
(c) What percentage of households spend between BWP 300 and BWP 499?
(d) Where do you expect the mean to lie relative to the median? Explain.

Exercise 5.2 (Box Plot Construction and Analysis). The marks (out of
100) scored by 15 students in a Statistics test are:
45, 52, 55, 58, 61, 64, 67, 70, 74, 78, 80, 83, 85, 90, 95
(a) Determine the five-number summary.
(b) Calculate the IQR and identify any outliers using the 1.5 × IQR rule.
(c) Sketch a box plot, clearly marking Q1 , Median, Q3 , whiskers, and any out-
liers.
(d) Describe what the box plot reveals about the spread and shape of the dis-
tribution.

Exercise 5.3 (Bar Chart vs Pie Chart). A survey of 200 Botho University
students asks which social media platform they use most frequently. The results
are:
WhatsApp: 90 Facebook: 50 Instagram: 35 TikTok: 15 Other: 10
(a) Calculate the relative frequency and pie chart angle for each category.
(b) Sketch a bar chart and a pie chart.
(c) Which chart communicates the data more clearly? In what scenario would
the other chart be preferred?
(d) If you added a sixth category “Twitter/X: 5” by moving 5 students from
“Other”, how would the pie chart angles change?

Exercise 5.4 (Choosing the Right Chart). For each dataset below, state
which chart type is most appropriate (histogram, bar chart, pie chart, box plot,
or ogive) and briefly justify your choice.

29
Descriptive Statistics — Lecture Notes Botho University

(a) The distribution of heights of 500 adult men (continuous, quantitative).

(b) The number of students enrolled in each of 6 university programmes.
(c) Comparing the spread of exam scores across three different classes.
(d) The market share (as %) of the four major mobile network operators in
Botswana.
(e) The distribution of daily temperatures in Gaborone over one year.
(f) Estimating what score a student must exceed to be in the top 20% of the
class.

Exercise 5.5 (Ogive). The table below shows the frequency distribution of the
heights (cm) of 50 plants in a nursery.

Height (cm) Frequency Upper Boundary

10 – 19 4 19.5
20 – 29 10 29.5
30 – 39 18 39.5
40 – 49 12 49.5
50 – 59 6 59.5

(a) Compute the cumulative frequencies and cumulative relative frequencies.

(b) Plot the ogive (use upper class boundaries on the horizontal axis).
(c) Use your ogive to estimate the median height.
(d) Use your ogive to estimate Q1 and Q3 .
(e) Estimate the 80th percentile (the height below which 80% of plants fall).

6 Shape of Distributions

6.1 Skewness
Skewness measures the degree and direction of asymmetry in a distribution.

Pearson’s Skewness Coefficient:

3(x̄ − Median)
Sk = (22)
s

Interpretation:
• Sk ≈ 0: symmetric distribution

30
Descriptive Statistics — Lecture Notes Botho University

• Sk > 0: right (positive) skew — long tail to the right; Mean > Median > Mode
• Sk < 0: left (negative) skew — long tail to the left; Mean < Median < Mode
• |Sk| > 1 indicates strong skew; |Sk| < 0.5 is approximately symmetric
Mode Mean=Med=Mode
Mode

Med
Med
Mean
Mean

value value value

LeftMean
(Negative) Skew
< Median < Mode
Symmetric
Sk ≈ 0
Right (Positive) Skew
Mode < Median < Mean

Example 6.1 — Skewness (House Prices).

From Example 3.5: x̄ = 288.75, Median = 222.5, s ≈ 212.5

3(288.75 − 222.5) 3 × 66.25 198.75

Sk = = = ≈ +0.94
212.5 212.5 212.5
Sk = +0.94 > 0: confirms right (positive) skew. The mansion at BWP 850k
stretches the distribution’s right tail.
Real-world context: Income distributions in most countries are right-skewed.
A small number of very high earners pull the mean above the median, which is
why median income better represents the “typical” worker.

6.2 Kurtosis
Kurtosis measures the peakedness of a distribution and the weight of its tails,
relative to the normal distribution (kurtosis = 3).

Excess kurtosis = κ − 3

The population kurtosis is defined as:

1 PN 4
i=1 (xi − µ)
κ= N (23)
σ4

In practice, most statistical software reports excess kurtosis (kurtosis − 3), so that a
normal distribution has excess kurtosis = 0. A positive excess kurtosis means heavier
tails than the normal; a negative excess kurtosis means lighter tails.

Kurtosis vs Variance: A common misconception is that kurtosis measures

“peakedness” in the same way variance measures spread. More precisely, kurtosis
is driven by the tails — datasets with the same variance can have very different
kurtosis if they differ in the frequency of extreme values. Two datasets can have
identical means, variances, and skewness but different kurtosis.

31
Descriptive Statistics — Lecture Notes Botho University

Type Kurtosis Excess Characteristics

Mesokurtic =3 =0 Normal bell curve (reference)

Leptokurtic >3 >0 Tall, narrow peak; heavy tails; more extreme values
Platykurtic <3 <0 Flat, broad peak; thin tails; values more evenly spread

Example 6.2 — Kurtosis in Finance.

Daily stock returns exhibit leptokurtosis (excess kurtosis > 0). Most trading days
produce small returns near zero (tall peak), but extreme crashes (e.g. −5% days)
occur more frequently than a normal distribution predicts (fat tails).
Implication: A risk model assuming normality would underestimate the proba-
bility of extreme losses. This is why finance professionals use Value-at-Risk (VaR)
and Conditional VaR (CVaR) instead of relying solely on standard deviation.
Data: {−5, −3, −2, −1, 0, 0, 0, 1, 1, 2, 3, 8} — mean ≈ 0.33, peaks at zero, but the
−5 and +8 values are much further from zero than a normal distribution would
suggest.

End-of-Chapter Exercises — Section 6

Exercise 6.1 (Skewness Calculation and Interpretation).
(a) A dataset has x̄ = 45, Median = 50, and s = 15. Calculate Pearson’s
skewness coefficient and describe the shape.
(b) For the following datasets, determine the skewness without calculation by
examining the values:
• {2, 3, 3, 4, 4, 4, 5, 5, 6}
• {10, 11, 12, 12, 13, 13, 13, 14, 50}
• {1, 8, 9, 9, 10, 10, 10, 11, 12}
(c) The annual incomes (BWP ’000) of employees at two firms are: Firm X:
x̄ = 55, Median= 48, s = 20; Firm Y: x̄ = 55, Median= 58, s = 14.
Calculate Sk for each and compare the income distributions.

Exercise 6.2 (Skewness and Central Tendency).

(a) For a right-skewed distribution with x̄ = 80 and Median = 72, use Pearson’s
formula to estimate the mode (assume s = 18).
(b) A dataset of exam scores has Mode = 70, Median = 65, Mean = 60. Describe
the skewness and draw a rough sketch of the distribution shape.
(c) In a left-skewed distribution, which measure is generally the largest: mean,
median, or mode? Explain why using the concept of the “pulling” effect of
the tail.

32
Descriptive Statistics — Lecture Notes Botho University

Exercise 6.3 (Kurtosis Interpretation).

(a) Explain the difference between a leptokurtic and a platykurtic distribution.
Give a real-world example of each.
(b) A quality control dataset has excess kurtosis = −0.8. Describe what this
means for the distribution of product measurements.
(c) Why is leptokurtosis (fat tails) particularly important in financial risk man-
agement? What might happen if a risk model assumes normality when the
true distribution is leptokurtic?

Exercise 6.4 (Comprehensive Shape Analysis).

The monthly profits (BWP ’000) of a small business over 12 months are:
−5, 8, 12, 15, 18, 20, 22, 25, 28, 30, 35, 80
(a) Calculate the mean, median, and mode.
(b) Calculate the standard deviation.
(c) Calculate Pearson’s skewness coefficient and interpret it.
(d) Which measure of central tendency best describes the “typical” monthly
profit? Justify.
(e) The business owner claims “we typically make BWP 24,000 profit per month.”
Is this claim well-supported by the data? What measure are they likely us-
ing, and is it appropriate?

7 Exercises with Full Solutions

7.1 Exercises — Central Tendency and Frequency

Q1. Daily sales (BWP): 120, 145, 98, 200, 155, 145, 175, 130, 145, 188. Find the
mean, median, and mode.
Solution:
• x̄ = 1501/10 = 150.1 BWP
• Ordered: 98, 120, 130, 145, 145, 145, 155, 175, 188, 200. Median = (145 +
145)/2 = 145 BWP
• Mode = 145 BWP (appears 3 times)
• Note: Mean > Median, mild right skew due to the BWP 200 sale.

Q2–Q3. 30 students: 8 scored in the 40s, 10 in the 50s, 7 in the 60s, 3 in the
70s, 2 in the 80s.

33
Descriptive Statistics — Lecture Notes Botho University

Frequency Table:

Class Midpoint Freq f Rel Freq Cum Freq

40–49 44.5 8 0.267 8
50–59 54.5 10 0.333 18
60–69 64.5 7 0.233 25
70–79 74.5 3 0.100 28
80–89 84.5 2 0.067 30
Total 30 1.000

Estimated Mean (midpoint method):

8(44.5) + 10(54.5) + 7(64.5) + 3(74.5) + 2(84.5) 356 + 545 + 451.5 + 223.5 + 169 1745
x̄ ≈ = = ≈
30 30 30

Median by interpolation: n/2 = 15. cf reaches 18 in class 50–59 (cf before =

8, f = 10, w = 10, L = 50):
15 − 8
M = 50 + × 10 = 50 + 7 = 57
10

Q4. Mean = 50, Median = 42. Shape and mode estimate.

Since Mean (50) > Median (42), the distribution is right-skewed.
Using Pearson’s approximate relationship for moderately skewed data: Mode ≈
3×Median −2×Mean = 3(42) − 2(50) = 126 − 100 = 26.

Q5. Weighted GPA (as in Example 3.2): x̄w ≈ 3.41

7.2 Exercises — Dispersion, Shape, and Visualisation

Q6. Data: 10, 11, 12, 13, 14, 15, 16, 17, 18, 18 (n=10)
• Range = 18 − 10 = 8
• Q1 = 12, Q3 = 17, IQR = 5
• x̄ = 144/10 = 14.4
• (xi − x̄)2 = (10 − 14.4)2 + · · · + (18 − 14.4)2 = 19.36 + 11.56 + 5.76 + 1.96 +
P
0.16 + 0.36 + 2.56 + 6.76 + 12.96 + 12.96 = 74.4
• s2 = 74.4/9 ≈ 8.27 s ≈ 2.88
• Fences: [12 − 7.5, 17 + 7.5] = [4.5, 24.5]. No outliers.

34
Descriptive Statistics — Lecture Notes Botho University

Q7. CV comparison:
3 7
CVA = × 100 = 25% CVB = × 100 = 35%
12 20
Fund A is relatively more consistent (lower CV per unit of return).

Q8. Mean=75, Median=80, Mode=85, s = 10

(a) Mean < Median < Mode ⇒ Left (negative) skew
3(75 − 80) −15
(b) Sk = = = −1.5 (strong negative skew)
10 10

Q9. Levels of measurement:

• Student ID number ⇒ Nominal (numbers as labels)
• Exam grade A/B/C ⇒ Ordinal (order meaningful, gaps unknown)
• Temperature in Kelvin ⇒ Ratio (true zero: 0 K = absolute zero)
• Number of siblings ⇒ Ratio (countable, true zero exists)

Q10. Box plot for Q6 data (see Example 4.1):

Five-number summary: Min=10, Q1 =12, Median=14.5, Q3 =17, Max=18. Whiskers
extend to 10 and 18; no outliers (both within [4.5, 24.5]). The box is slightly asym-
metric — median is closer to Q1 than Q3 , indicating mild right skew.

8 Chapter Review: Mixed Exercises

These exercises integrate concepts from all sections. They are suitable for exam revi-
sion.
Review Exercise 1 (Comprehensive).
The table below shows the distribution of the number of hours per week that 60
adults in Gaborone spend on physical exercise.

35
Descriptive Statistics — Lecture Notes Botho University

Hours per Week Frequency Cumulative Frequency

0–1 10 10
2–3 18 28
4–5 20 48
6–7 8 56
8–9 4 60

(a) Estimate the mean number of exercise hours using the midpoint method.
(b) Use interpolation to find the median.
(c) Identify the modal class. Estimate the mode.
(d) Calculate the relative frequency for each class.
(e) Based on your answers to (a)–(c), describe the shape of the distribution.
(f) Calculate Pearson’s skewness coefficient. Does it confirm your shape de-
scription?
(g) A health authority states that adults should exercise at least 4 hours per
week. What percentage of this sample meets the target?

Review Exercise 2 (Data Analysis Case Study).

A supermarket records the number of items purchased per customer for 15 trans-
actions:
3, 7, 12, 5, 18, 9, 7, 22, 6, 14, 7, 4, 35, 11, 8
(a) Calculate the mean, median, and mode.
(b) Determine the five-number summary and IQR.
(c) Identify any outliers using the 1.5 × IQR rule. If outliers are found, recal-
culate the mean and median without them and comment on the change.
(d) Calculate the sample standard deviation.
(e) Calculate the coefficient of variation and interpret it.
(f) Calculate Pearson’s skewness. Is the distribution symmetric, left-skewed, or
right-skewed?
(g) Sketch a box plot for the original data, showing any outliers as separate
points.

Review Exercise 3 (Multiple Choice Concepts). For each question, choose

the best answer and explain your reasoning.
1. Which measure of central tendency is most appropriate for ordinal data?

36
Descriptive Statistics — Lecture Notes Botho University

(a) Mean (b) Median (c) Mode (d) All three are appropriate
2. A dataset has variance s2 = 64. Its standard deviation is:
(a) 64 (b) 32 (c) 8 (d) 4,096
3. Two distributions have the same mean and standard deviation but different
kurtosis. This means:
(a) They are identical distributions
(b) They differ in the weight of their tails
(c) One is skewed and one is not
(d) They have different ranges
4. In a right-skewed distribution, which ordering is correct?
(a) Mode > Median > Mean
(b) Mean > Median > Mode
(c) Median > Mean > Mode
(d) Mean = Median = Mode
5. The coefficient of variation is most useful when:
(a) Comparing the spread of datasets with the same units
(b) Comparing the spread of datasets with different units or very different
means
(c) The data contains outliers
(d) The distribution is symmetric

Review Exercise 4 (Short Answer). Answer each part in 2–4 sentences.

(a) Explain why frequency distribution tables are useful when working with
large datasets.
(b) A data analyst reports the mean income of a neighbourhood as BWP 35,000.
A local newspaper reports the median income as BWP 22,000. Which figure
should citizens use to understand the typical resident’s income? Why are
the two figures so different?
(c) Describe a real-world situation where a high coefficient of variation (e.g.,
CV > 80%) would be expected and one where a low CV (e.g., CV < 10%)
would be expected.
(d) A student says: “I increased every score in my dataset by 10 marks. This
will increase the standard deviation.” Is the student correct? Explain.
(e) Describe the difference between a bar chart and a histogram. When should
each be used?

37
Descriptive Statistics — Lecture Notes Botho University

9 Worked Solutions to Selected End-of-Chapter Ex-

ercises

Solutions — Section 1
Exercise 1.1 — Solutions.
(a) Number of goals: Quantitative, discrete (countable whole numbers).
(b) Time in minutes: Quantitative, continuous (can take any value in a
range).
(c) Colour of car: Qualitative (category label, no numeric meaning).
(d) Monthly salary (BWP): Quantitative, continuous (any positive real value).
(e) Number of ATM transactions: Quantitative, discrete (countable whole
numbers).
(f) Satisfaction rating Poor/Fair/Good/Excellent: Qualitative (ordered cate-
gories with no fixed numeric gap).

Exercise 1.2 — Solutions.

(a) Year born: Interval. Years are ordered with equal gaps, but “year 0” is a
calendar convention, not an absolute zero.
(b) Blood pressure (mmHg): Ratio. Has a true zero (no pressure = 0 mmHg);
ratios are meaningful.
(c) Military rank: Ordinal. Order is meaningful (Sergeant outranks Corporal)
but the “distance” between ranks is not quantified.
(d) Country of birth: Nominal. Categories with no natural order.
(e) Temperature (°C): Interval. Equal degree gaps, but 0°C is not “no tem-
perature” — it is the freezing point of water.
(f) Distance driven (km): Ratio. True zero (no distance driven = 0 km); ratios
meaningful (200 km is twice 100 km).

Exercise 1.4 — Solution.

(a) This involves descriptive statistics — it summarises existing (sample)
data.
(b) The CEO’s BWP 500,000 bonus is an extreme outlier that pulls the mean far
above what a typical employee receives. If there are, say, 49 other employees
and each earns a BWP 2,000 bonus, the mean is (49 × 2000 + 500000)/50 =
BWP 11,960 — yet nobody except the CEO earns near that amount.

38
Descriptive Statistics — Lecture Notes Botho University

(c) The median would better represent the typical employee’s bonus, as it is
not affected by the CEO’s extreme value.

Solutions — Section 2
Exercise 2.1 — Solution.
Sorted ages: 23, 23, 26, 27, 29, 30, 31, 31, 34, 37, 38, 39, 41, 44, 45, 48, 52, 55
√
(a) Range = 55 − 23 = 32. k ≈ 18 ≈ 4.2 → use 5 classes. Width = 32/5 =
6.4 → round up to 7 (or use width 10 for cleaner boundaries).
(b) Using width = 7, starting at 22:

Class Freq Rel Freq Cum Freq Cum Rel Freq

22–28 4 0.222 4 0.222
29–35 5 0.278 9 0.500
36–42 4 0.222 13 0.722
43–49 3 0.167 16 0.889
50–56 2 0.111 18 1.000
Total 18 1.000

(c) Employees younger than 40: the class 36–42 contains employees up to age
42, so we count all in classes up to 36–42 who are below 40. From the raw
sorted data, ages below 40 are: 23,23,26,27,29,30,31,31,34,37,38,39 — that is 12
employees. 12/18 × 100 = 66.7%.
(d) Modal class: 29–35 (frequency 5, the highest).

Exercise 2.2 — Solution (key parts).

(a) Relative frequencies: 8/50 = 0.16; 14/50 = 0.28; 16/50 = 0.32; 9/50 = 0.18;
3/50 = 0.06. Cumulative relative frequencies: 0.16, 0.44, 0.76, 0.94, 1.00.
(b) Subscribers using 5–14 GB: (14 + 16)/50 = 30/50 = 60%.
(c) Only 12% fall above: we need the 88th percentile. Cumulative relative fre-
quency reaches 0.76 at the end of 10–14 GB class. 0.76 + 0.18 = 0.94 at end of
15–19 GB class. Since 88% lies within the 15–19 class, use interpolation:
0.88 − 0.76 0.12
P88 = 15 + × 5 = 15 + × 5 = 15 + 3.33 ≈ 18.3 GB
0.18 0.18

(d) The histogram peaks in the 10–14 GB class, with fewer subscribers at both
extremes. Shape: approximately symmetric, slightly right-skewed (the 20–24
class is smaller than the 0–4 class).

Solutions — Section 3

39
Descriptive Statistics — Lecture Notes Botho University

Exercise 3.1 — Solution.

Original data: 14, 18, 22, 9, 31, 18, 25, 18, 12, 27, 20, 15
P
(a) x = 229. x̄ = 229/12 ≈ 19.1.
Sorted: 9, 12, 14, 15, 18, 18, 18, 20, 22, 25, 27, 31.
Median = (18 + 18)/2 = 18. Mode = 18 (appears 3 times).
P
(b) Replace 31 with 13: x = 229 − 31 + 13 = 211. x̄ = 211/12 ≈ 17.6.
Sorted: 9, 12, 13, 14, 15, 18, 18, 18, 20, 22, 25, 27.
Median = (18 + 18)/2 = 18. Mode = 18 (unchanged).
(c) The mean changed the most (dropped by about 1.5), while the median and
mode were unchanged. This demonstrates that the mean is sensitive to extreme
values, while the median and mode are robust.

Exercise 3.2 — Solution.

(a) Grouped mean:

3(2) + 8(7) + 14(12) + 10(17) + 5(22) 6 + 56 + 168 + 170 + 110 510

x̄ ≈ = = = 12.75 hours
40 40 40

(b) Grouped median: n/2 = 20. Cumulative frequency: 3, 11, 25 (reaches 20

in class 10–14). L = 10, cf = 11, f = 14, w = 5:
20 − 11 9
M = 10 + × 5 = 10 + × 5 = 10 + 3.21 = 13.21 hours
14 14

(c) Modal class: 10–14 (frequency 14).

(d) Students meeting target (≥ 12 hours): from cumulative frequency, 14 students
are in 10–14, 10 in 15–19, 5 in 20–24. Those in 10–14 who score ≥ 12 requires
interpolation: (14 − 12)/5 × 14 ≈ 5.6 are below 12 in that class, so 14 − 5.6 = 8.4
are ≥ 12 within that class. Total ≥ 12: 8.4 + 10 + 5 = 23.4/40 ≈ 58.5%.
(Alternatively, using the ogive: cumulative frequency at 12 ≈ 11 + 0 = 11, so
40 − 11 = 29 students score ≥ 12 ⇒ 29/40 = 72.5%, using the class boundary.)

Exercise 3.3 — Solution.

(a) Weighted mean:

5(72) + 2(85) + 3(68) + 4(78) 360 + 170 + 204 + 312 1046

x̄w = = = = 74.7%
5+2+3+4 14 14

(b) Simple mean: (72 + 85 + 68 + 78)/4 = 303/4 = 75.75%.

(c) The unweighted mean is slightly higher. Communication Skills (85%) has a
high score but only 2 credits; since the weighted mean gives it less influence, the

40
Descriptive Statistics — Lecture Notes Botho University

lower-scoring Business Mathematics (72%, 5 credits) drags the weighted mean

down.
(d) Current total weighted points = 1046. New module: 3 credits, score x. Need
weighted average ≥ 75%:
1046 + 3x
≥ 75 ⇒ 1046 + 3x ≥ 1275 ⇒ 3x ≥ 229 ⇒ x ≥ 76.3%
14 + 3
The student must score at least 76.3% in the new module.

Exercise 3.4 — Solution.

(a) Current total salary = 8 × 18500 = BWP 148,000. New total = 148000 +
22000 = 170000. New mean = 170000/9 = BWP 18,889.
(b) New total = 148000+14000 = 162000. New mean = 162000/9 = BWP 18,000.
(c) Combined mean:

5(16000) + 3(19500) 80000 + 58500 138500

x̄combined = = = = BWP 17,312.50
5+3 8 8

Solutions — Section 4
Exercise 4.1 — Solution.
Branch A (sorted): 4, 5, 5, 6, 6, 7, 7, 8, 8, 9. x̄A = 65/10 = 6.5.
Q1 = 5, Q3 = 8, IQRA = 3. Fences: [0.5, 12.5]. No outliers.
(x − x̄)2 = (4 − 6.5)2 + · · · + (9 − 6.5)2 = 6.25 + 2.25 + 2.25 + 0.25 + 0.25 +
P
0.25 + 0.25 + 2.25 + 2.25 + 6.25 = 22.5
s2A = 22.5/9 = 2.5, sA ≈ 1.58, CVA = (1.58/6.5) × 100 = 24.3%.
Branch B (sorted): 1, 2, 2, 3, 4, 6, 9, 10, 11, 12. x̄B = 60/10 = 6.0.
Q1 = 2, Q3 = 10, IQRB = 8. Fences: [−10, 22]. No outliers.
(x − 6)2 = 25 + 16 + 16 + 9 + 4 + 0 + 9 + 16 + 25 + 36 = 156
P
s2B = 156/9 ≈ 17.33, sB ≈ 4.16, CVB = (4.16/6.0) × 100 = 69.3%.
(d) Branch A is more consistent: lower SD (1.58 vs 4.16), lower IQR (3 vs 8),
and much lower CV (24.3% vs 69.3%). Both means are similar, but Branch B’s
service times vary wildly from 1 to 12 minutes.

Exercise 4.2 — Solution.

(a) z = (78 − 63)/9 = 15/9 ≈ +1.67. This student scored 1.67 standard devi-
ations above the class mean — a strong performance, approximately in the top
5%.
(b) z = (48 − 63)/9 = −15/9 ≈ −1.67. This student scored 1.67 SDs below the
mean — symmetrically poor.
(c) x̄ ± 2s = 63 ± 18 = [45, 81]. About 95% of students scored between 45 and

41
Descriptive Statistics — Lecture Notes Botho University

81.
(d) x = x̄ + z · s = 63 + (−2.5)(9) = 63 − 22.5 = 40.5.
(e) Thabo: z = (72 − 65)/10 = 0.70. Kefilwe: z = (80 − 75)/8 = 0.625. Thabo
has a higher z-score, so he performed better relative to his class.

Exercise 4.4 — Solution (Computational Shortcut Verification).

Data: 3, 7, 7, 9, 10, 12, 15. n = 7.
P P 2
xi = 63. xi = 9 + 49 + 49 + 81 + 100 + 144 + 225 = 657.
Shortcut:
657 − 632 /7 657 − 3969/7 657 − 567 90
s2 = = = = = 15
6 6 6 6

Verification using definition: x̄ = 63/7 = 9.

X
(xi − 9)2 = 36 + 4 + 4 + 0 + 1 + 9 + 36 = 90. s2 = 90/6 = 15✓

√
s= 15 ≈ 3.87.

Exercise 4.6 — Solution.

(a) Flat raise of BWP 800: ȳ = 14200 + 800 = BWP 15,000. sy = BWP 3,100
(unchanged — adding a constant does not change spread).
(b) 10% raise (b = 1.10): ȳ = 1.10 × 14200 = BWP 15,620. sy = 1.10 × 3100 =
BWP 3,410.
(c) The percentage raise produces larger spread (BWP 3,410 vs BWP 3,100).
Intuitively, a flat raise shifts everyone equally — the gaps between salaries stay
identical. A percentage raise widens the gap between high and low earners because
higher salaries receive larger absolute increases.
(d) In USD: ȳ = 14200/13.5 ≈ 1051.9 USD. sy = 3100/13.5 ≈ 229.6 USD.
CVUSD = (229.6/1051.9) × 100 ≈ 21.8%.
CVBWP = (3100/14200) × 100 ≈ 21.8%.
The CV is identical — dividing by a constant scales both mean and SD equally,
leaving their ratio unchanged.

Exercise 4.7 — Solution.

Data (n = 16): 85, 102, 115, 128, 134, 141, 155, 163, 170, 182, 195, 208, 224, 237,
260, 315
(a) Q1 : L = (25/100) × 16 = 4. Q1 = (x4 + x5 )/2 = (128 + 134)/2 = 131.
Median: L = 8. Q2 = (x8 + x9 )/2 = (163 + 170)/2 = 166.5.
Q3 : L = 12. Q3 = (x12 + x13 )/2 = (208 + 224)/2 = 216.

42
Descriptive Statistics — Lecture Notes Botho University

(b) IQR = 216 − 131 = 85. Fences: [131 − 127.5, 216 + 127.5] = [−96.5, 343.5].
All values within range ⇒ no outliers.
(c) P30 : L = 0.30 × 16 = 4.8 → round up to 5. P30 = x5 = 134.
P70 : L = 0.70 × 16 = 11.2 → round up to 12. P70 = x12 = 208.
P90 : L = 0.90 × 16 = 14.4 → round up to 15. P90 = x15 = 260.
(d) P90 = BWP 260 means that 90% of the sampled households pay less than
BWP 260 per month for electricity, and only 10% pay more.
(e) The middle 50% of bills fall between Q1 = BWP 131 and Q3 = BWP 216.
This range is the interquartile range (IQR).

Solutions — Section 5
Exercise 5.0 — Solution (Stem-and-Leaf ).
Sorted ages: 23, 28, 29, 31, 33, 34, 36, 38, 42, 44, 45, 47, 48, 51, 52, 55, 61, 67
(a)
Stem Leaves
2 389
3 13468
4 24578
5 125
6 17
(b) n = 18. Median = average of 9th and 10th values = (42 + 44)/2 = 43.
Q1 = median of lower 9: 4th value = 34.
Q3 = median of upper 9: 14th value = 52 (counting from the 10th observation
up: 44,45,47,48,51,52,55,61,67 — 6th of these = 52).
(c) The distribution is approximately symmetric to mildly right-skewed. The
30s and 40s rows each have 5 leaves; the tails at 20s (3 leaves) and 50s–60s (5
leaves) are roughly balanced.
(d) Back-to-back plot:
Group 1 (right to left) Stem Group 2
1 9
9 8 3 2 5
864 3 1 3 058
875 4 2 4 2
5 2 1 5 08
7 1 6 3
7 1
Group 2 has a wider age range (19–71) with participants at both younger and
older extremes. Group 1 is more concentrated in the 30s–50s. Group 2’s median
appears lower (around 40) vs Group 1’s median of 43.

43
Descriptive Statistics — Lecture Notes Botho University

Exercise 5.2 — Solution (Box Plot).

Data (n = 15): 45, 52, 55, 58, 61, 64, 67, 70, 74, 78, 80, 83, 85, 90, 95
(a) Five-number summary: Min=45, Q1 = 58, Median=70, Q3 = 83, Max=95.
(Q1 : 4th value = 58; Median: 8th value = 70; Q3 : 12th value = 83)
(b) IQR = 83 − 58 = 25. Fences: [58 − 37.5, 83 + 37.5] = [20.5, 120.5]. All values
within range ⇒ no outliers.
(c) Box: from 58 to 83, with median line at 70. Whiskers from 45 (left) to 95
(right).
(d) The box is roughly symmetric (median at 70 is close to the midpoint of 58
and 83 = 70.5). The distribution covers a wide range (45–95). No extreme values
stand out; spread is moderate and fairly uniform.

Solutions — Section 6
Exercise 6.1 — Solution.
(a) Sk = 3(45 − 50)/15 = 3(−5)/15 = −1.0. Strong left (negative) skew —
the long tail extends to the left.
(b)
• {2, 3, 3, 4, 4, 4, 5, 5, 6}: mode=4, data balanced on either side. Approxi-
mately symmetric.
• {10, 11, 12, 12, 13, 13, 13, 14, 50}: the value 50 is a high outlier pulling the
mean right. Right-skewed.
• {1, 8, 9, 9, 10, 10, 10, 11, 12}: the value 1 is a low outlier pulling the mean
left. Left-skewed.
(c) Firm X: Sk = 3(55 − 48)/20 = 21/20 = +1.05 (strong positive/right skew —
a few very high earners).
Firm Y: Sk = 3(55 − 58)/14 = −9/14 = −0.64 (moderate negative/left skew —
a few low earners drag the mean below the median).
Both firms have the same mean salary but opposite skew: Firm X’s distribution
is pulled up by high earners; Firm Y’s is pulled down by low earners.

Exercise 6.4 — Solution.

Data: −5, 8, 12, 15, 18, 20, 22, 25, 28, 30, 35, 80. n = 12.
P
(a) x = 288. x̄ = 288/12 = 24 BWP ’000.
Median (avg of 6th and 7th): (20 + 22)/2 = 21 BWP ’000.
Mode: none (all values unique).
(b) x̄ = 24.
X
(x − 24)2 = 841 + 256 + 144 + 81 + 36 + 16 + 4 + 1 + 16 + 36 + 121 + 3136 = 4688

44
Descriptive Statistics — Lecture Notes Botho University

s2 = 4688/11 ≈ 426.2. s ≈ 20.6 BWP ’000.

(c) Sk = 3(24 − 21)/20.6 = 9/20.6 ≈ +0.44. Mild positive (right) skew — the
BWP 80k month stretches the right tail.
(d) The median (BWP 21,000) best represents the typical month; the outlier
month of BWP 80k inflates the mean by BWP 3,000 above a value that 10 of 12
months do not reach.
(e) The owner is likely using the mean (BWP 24,000 = x̄). This is misleading
because it is driven upward by a single exceptional month. The median (BWP
21,000) is more representative of a typical trading month.

10 Practice Examination

Descriptive Statistics — Practice Examination

Botho University — Quantitative Methods
Time allowed: 2 hours Total marks: 100
Answer ALL questions. Show all working clearly. Calculators permitted.

Question 1 — Data Classification and Levels of Measurement [15

marks]
(a) For each variable below, state the type (qualitative/quantitative; discrete/continuous
if quantitative) and the level of measurement. Give a one-sentence justification
for each level. [10]
(i) Annual rainfall in Botswana (mm)
(ii) Class of airline ticket: Economy / Business / First
(iii) Employee ID number at a company
(iv) Number of text messages sent per day
(v) Customer satisfaction score on a scale of 1–10
(b) A researcher collects data on: the gender of respondents, their monthly in-
come (BWP), and their level of education (no schooling / primary / secondary
/ tertiary). State the most appropriate measure of central tendency for each
variable and briefly justify. [5]

Question 2 — Frequency Distributions and Visualisation [25 marks]

The following data represents the time (minutes) taken by 30 call centre agents
to resolve customer queries:
8, 12, 5, 18, 22, 9, 14, 27, 6, 11, 16, 24, 10, 19, 7,

45
Descriptive Statistics — Lecture Notes Botho University

13, 25, 8, 21, 15, 9, 17, 23, 11, 6, 20, 14, 28, 10, 16
(a) Determine a suitable number of classes and class width. Construct a com-
plete frequency distribution table showing: class intervals, midpoints, frequency,
relative frequency, cumulative frequency, and cumulative relative frequency. [8]
(b) Construct a histogram for the data. Label both axes clearly. [4]
(c) Construct a stem-and-leaf plot for the data. [4]
(d) Compute the cumulative frequencies and draw an ogive. Use your ogive to
estimate the median and the 75th percentile. [5]
(e) Based on your histogram, describe the shape of the distribution. [4]

Question 3 — Measures of Central Tendency [20 marks]

(a) Using your frequency table from Question 2:
(i) Estimate the mean resolution time using the midpoint method. [3]
(ii) Find the median by interpolation. Compare your answer to the ogive esti-
mate. [4]
(iii) Identify the modal class and estimate the mode. [3]
(b) A company employs three departments. Department P has 10 employees with
a mean monthly salary of BWP 12,000. Department Q has 6 employees with a
mean monthly salary of BWP 18,500. Department R has 4 employees with a mean
monthly salary of BWP 25,000.
(i) Calculate the combined mean monthly salary for all 20 employees. [4]
(ii) An external consultant charges BWP 350 per hour. She works 5 hours per
department meeting (3 meetings). What was her total fee, and what was
her average (mean) fee per meeting? [3]
(iii) Comment on which measure of central tendency is most appropriate to rep-
resent a “typical” salary in this company, given that the three departments
differ significantly in pay level. [3]

Question 4 — Measures of Dispersion [25 marks]

(a) Using the raw call centre data from Question 2:
(i) Calculate the mean and sample standard deviation. Show all working clearly.
[6]
(ii) Find the five-number summary and IQR. Check for outliers. [5]
(iii) Calculate the coefficient of variation. Interpret your result. [3]
(b) A competing call centre has a mean resolution time of 17.5 minutes and a
standard deviation of 9.8 minutes.

46
Descriptive Statistics — Lecture Notes Botho University

(i) Compare the relative variability of the two call centres using the CV. Which
centre is more consistent? [3]
(ii) An agent at the competing centre resolved a query in 5 minutes. Calculate
and interpret her z-score. [3]
(iii) Using the Empirical Rule, between what two times do approximately 95%
of the competing centre’s resolution times fall? State any assumption you
are making. [3]

Question 5 — Shape of Distributions [15 marks]

(a) The monthly net profits (BWP ’000) of a retail shop over the past year are:
14, 18, 22, 19, 25, 17, 21, 16, 23, 28, 20, 24
(i) Calculate the mean, median, and standard deviation. [4]
(ii) Calculate Pearson’s skewness coefficient and interpret the result. Is the
distribution left-skewed, right-skewed, or approximately symmetric? [3]
(b) The table below summarises the distribution of waiting times (minutes) at a
government office.

Waiting Time (min) Frequency Midpoint

0 – 9 5 4.5
10 – 19 12 14.5
20 – 29 20 24.5
30 – 39 8 34.5
40 – 49 3 44.5
50 – 59 2 54.5

(i) Estimate the mean and median waiting time. [4]

(ii) Calculate Pearson’s skewness and identify the direction of skew. [2]
(iii) A government target states that 80% of customers should be served within
30 minutes. Does this sample suggest the office is meeting this target?
Justify using the cumulative frequency. [2]

— End of Practice Examination —

11 Key Formulas Summary

47
Descriptive Statistics — Lecture Notes Botho University

Measure Formula
P
xi
Sample mean x̄ =
n
P
wi xi
Weighted mean x̄w = P
P wi
fi mi
Mean from grouped data x̄ ≈
n
n/2 − cf
Median (grouped) M =L+ ×w
f
d1
Mode (grouped) Mode ≈ L + ×w
d1 + d 2
Range R = Max − Min

IQR IQR = Q3 − Q1
(xi − x̄)2
P
2
Sample variance s =
P n2− 1 P 2
xi − ( xi ) /n
Shortcut variance s2 =
n−1
√
Standard deviation s = s2
x − x̄
Z-score z=
s
s
Coefficient of variation CV = × 100%
x̄
Linear transformation ȳ = a + bx̄; sy = |b| · sx
3(x̄ − Median)
Pearson’s skewness Sk =
1
P s 4
(xi − µ)
Kurtosis (population) κ= N
σ4
Outlier fences Q1 − 1.5 × IQR ; Q3 + 1.5 × IQR

Relative frequency rfi = fi /n

fi
Pie chart angle θi = × 360
n

Learning Outcomes

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

1. Distinguish between a population and a sample, and between a parameter and a
statistic.
2. Classify variables as qualitative or quantitative, discrete or continuous, and identify
the level of measurement (nominal, ordinal, interval, ratio).

48
Descriptive Statistics — Lecture Notes Botho University

3. Construct and interpret frequency, relative frequency, and cumulative frequency

tables, correctly applying class limits, class boundaries, and midpoints.
4. Draw and interpret an ogive, and use it to read off percentiles and quartiles.
5. Calculate and interpret the mean, median, and mode — for both raw and grouped
data (including the midpoint method for the mean and interpolation for the median
and mode) — and choose the most appropriate measure for a given context.
6. Compute the range, IQR, variance (using both the definition and the computational
shortcut), standard deviation, coefficient of variation, and z-score.
7. Apply the rules for how linear transformations (adding/multiplying a constant)
affect the mean, standard deviation, and variance.
8. Detect potential outliers using the 1.5 × IQR rule and construct box plots.
9. Describe the shape of a distribution using skewness (Pearson’s coefficient) and
kurtosis, and interpret the relationship between mean, median, and mode under
skew.
10. Construct and interpret histograms, ogives, box plots, bar charts, and pie charts,
and select the appropriate chart type for a given dataset and purpose.

49