Statistics — Study Notes

Measures of Central Tendency & Dispersion · Correlation & Regression · Probability ·

Worked Problems

Compiled and typed from handwritten class notes (May–June 2026). Organised by topic for
easier reference and revision.

Contents
• 1. Measures of Central Location / Tendency — basic definitions
• 2. Mean Deviation (M.D.)
• 3. Quartile Deviation (Q.D.)
• 4. Median and Mode from Grouped Data
• 5. Cumulative Frequency & the Median Formula
• 6. Worked Problems — Corrected Mean/Median, Standard Deviation
• 7. AM, GM, HM — Relations and Combined (Composite) Mean
• 8. Correlation & Regression — Concepts and Scatter Diagrams
• 9. Karl Pearson's Correlation Coefficient
• 10. Regression Coefficients
• 11. Properties of the Correlation Coefficient
• 12. Worked Example — Correlation via Step-Deviation (u, v)
• 13. Probability — Theorem of Compound Probability
• 14. Spearman's Rank Correlation Coefficient
1. Measures of Central Location / Tendency
06/05/2026

Variable vs Constant
Variable: a quantity that can take different values — e.g. x, y, z, u, v ...
Constant: a quantity with a fixed value — e.g. a, b, c, d ...

Class Boundary and Class Limit

• C/B — Class Boundary
• C/L — Class Limit
• Frequency (f) → the number of occurrences of a value / class

Converting Class Limits to Class Boundaries

Correction factor = ½ = 0.5 (added to the upper limit, subtracted from the lower limit) when
class limits are not continuous.

C/L (as given) C/B (continuous) Class Mark (mid-value)

10 – 15 9.5 – 15.5* (10+15)/2 = 12.5

15 – 20 — (15+20)/2 = 17.5

20 – 29 19.5 – 29.5 —

30 – 39 29.5 – 39.5 —

40 – 50 39.5 – 50.5 —
*Class boundaries are used when classes must be made continuous for grouping/graphing.

Frequency Table Notation (single-value class)

Observation Frequency f·x

x₁ f₁ f₁x₁

x₂ f₂ f₂x₂

x₃ f₃ f₃x₃

⋮ ⋮ ⋮

xₙ fₙ fₙxₙ
Σ fᵢ = N (total number of observations)

Σxᵢ (i = 1 to 10) can be split as Σxᵢ (i = 1 to 6) + Σxᵢ (i = 7 to 10) — summation can

always be broken into partial sums.
2. Mean Deviation (M.D.)
Also called Mean Absolute Deviation (MAD).

Formula
For simple (ungrouped) observations:
MD = (1/n) Σ |xᵢ − x̄|, where x̄ = Σx / n

For grouped observations:

MD = (1/N) Σ fᵢ |xᵢ − x̄|, where x̄ = Σfᵢxᵢ / N (xᵢ = class mid-value)

Note: the vertical bars denote the modulus (absolute value) — deviations are taken as
positive regardless of sign, since on expansion Σ|xᵢ − x̄| ≠ |Σxᵢ − nx̄| in general (the modulus
must be applied term-by-term).

Worked Example — Mean Deviation from a Grouped Frequency

Table
Step-deviation method used: y = (xᵢ − 45)/10 (origin = 45, scale = 10).

C/B f c.f. Mid-value (xᵢ) y = (xᵢ−45)/10 f·y |xᵢ − x̄| f·|xᵢ − x̄|

10–20 5 5 15 −3 −15 31 155

20–30 7 12 25 −2 −14 21 147

30–40 18 30 35 −1 −18 11 198

40–50 31 61 45 0 0 1 31

50–60 24 85 55 1 24 9 216

60–70 12 97 65 2 24 19 228

70–80 3 100 75 3 9 29 87

Total N = 100 Σfy = 10 Σf|xᵢ−x̄| = 1062

ȳ = Σfy / N = 10/100 = 0.1

x̄ = 45 + 10 × 0.1 = 46

MD = Σfᵢ|xᵢ − x̄| / N = 1062 / 100 = 10.62

3. Quartile Deviation (Q.D.)
15/05/2026

Q.D. = (Q₃ − Q₁) / 2

Q₁ = l₁ + [ (N/4 − F) / f₁ ] × C

Q₃ = l₃ + [ (3N/4 − F) / f₃ ] × C

where l = lower class boundary of the quartile class, F = cumulative frequency before the
quartile class, f = frequency of the quartile class, C = class width.

Worked Example (N = 100)

3N/4 = 75 → Q₃ = 50 + [(75 − 61)/24] × 10 = 50 + (14/24)×10 = 50 + 5.83 = 55.83

N/4 = 25 → Q₁ = 30 + [(25 − 12)/18] × 10 = 30 + (130/18) = 30 + 7.22 = 37.22

Q.D. = (55.83 − 37.22)/2 = 18.61/2 = 9.305

Coefficient of Quartile Deviation

Coefficient of Q.D. = (Q.D. / Q₂) × 100

where Q₂ is the 2nd quartile (= median, Mc).

Q₂ = Mc = l + [(N/2 − F)/f] × C = 40 + [(50−30)/31] × 10 = 40 + 6.45 = 46.45

Coefficient of Q.D. = (9.305 / 46.45) × 100 = 20.03

4. Median and Mode from Grouped Data
13/05/2026

Class Frequency (f) c.f.

10 – 20 5 5

20 – 30 7 12

30 – 40 18 30

40 – 50 (median class) 31 61

50 – 60 24 85

60 – 70 12 97

70 – 80 3 100

Total N = Σf = 100

Step 1 — Locate N/2

N/2 = 100/2 = 50

Step 2 — Median Formula (grouped data)

Mc = l₁ + [ (N/2 − F) / fₘ ] × C

where l₁ = lower boundary of median class, F = c.f. before the median class, fₘ = frequency
of median class, C = class width.
Mc = 40 + [(50 − 30)/31] × 10 = 40 + (20/31)×10 = 40 + 200/31 = 40 + 6.45 = 46.45

Mode
Definition: the value of the variable corresponding to the highest (maximum) frequency,
found after arranging the data in ascending order.
Example: Raw data (n = 12): 11, 7, 13, 8, 4, 2, 4, 8, 6, 12, 15, 8
Arranged ascending: 2, 4, 4, 6, 7, 8, 8, 8, 11, 12, 13, 15
The value 8 occurs 3 times — the highest frequency among all values.
Mode (Mo) = 8
5. Cumulative Frequency and the Median Formula
Cumulative frequency (c.f.) of a class = sum of the frequency of that class and all preceding
classes.

• CF< (less-than type): running total from the lowest class upward — this is the ordinary
c.f.
• CF> (more-than type): running total from the highest class downward

Class f c.f. (less-than type)

x₀ – x₁ f₁ f₁

x₁ – x₂ f₂ f₁ + f₂

x₂ – x₃ f₃ f₁ + f₂ + f₃

⋮ ⋮ ⋮

xₙ₋₁ – xₙ fₙ Σfᵢ = N

Construction Steps
• Keep the class in Class Boundary (C/B) form; if given as Class Limits (C/L), convert
first.
• Build the cumulative (less-than type) frequency from the ordinary frequency.
• Apply the median formula.
Mc = l₁ + [ (N/2 − F) / fₘ ] × C

l₁ = lower class boundary of the median class | N/2 = half the total frequency | F = c.f.
below l₁ (preceding the median class) | fₘ = ordinary frequency of the median class | C =
common class width.
Definition of Median: Mc is the value of the variable corresponding to a cumulative frequency
of N/2. It is also called the 2nd Quartile.
6. Worked Problems
25/05/2026

Problem 1 — Corrected Mean and Median

The mean and median of 100 items are 50 and 50 respectively. It was later found that the
value of the largest item was wrongly recorded as 100 instead of 110. Find the corrected
mean and median.

Solution
Σxᵢ = mean × n = 50 × 100 = 5000

Corrected Σxᵢ = 5000 − 100 + 110 = 5010

Corrected mean = 5010 / 100 = 50.1

Since n = 100 (even), the median is the average of the 50th and 51st observations in the
ascending arrangement. The item that was corrected is the largest item (last position), so
the 50th and 51st observations are unaffected.
Corrected median = 50 (unchanged)

Problem 2 — Standard Deviation of 5, 5, 5, 7, 7, 7

n = 6, Σx = 5+5+5+7+7+7 = 36, Σx² = 25+25+25+49+49+49 = 222
V(x) = σ² = Σx²/n − (Σx/n)² = 222/6 − (36/6)² = 37 − 36 = 1

SD = √V(x) = √1 = 1

Problem 3 — Mean from Algebraic Sum of Deviations

If the algebraic sum of deviations of 20 observations, measured from 30, is 2, find the mean
of the observations.

Solution
Σ(xᵢ − 30) = 2 (i = 1 to 20)

⇒ Σxᵢ − 20×30 = 2 ⇒ Σxᵢ = 2 + 600 = 602

x̄ = Σxᵢ / n = 602/20 = 30.1

Problem 4 — SD of First 10 Natural Numbers

(each with equal/unit frequency)
Σx = 1+2+...+10 = n(n+1)/2 = 10×11/2 = 55
Σx² = n(n+1)(2n+1)/6 = 10×11×21/6 = 385
V(x) = σ² = Σx²/n − (Σx/n)² = 385/10 − (55/10)² = 38.5 − 30.25 = 8.25

SD = √8.25 = 2.87 (approx.)

Problem 5 — Standard Deviations of Two Data Sets

Set n Σx Σx² Variance SD

x₁ (5 obs.) n₁ = 5 40 410 82/n₁ − (40/5)² = 82 − 64 = 18√18 = 4.24

x₂ (3 obs.) n₂ = 3 24 264 264/3 − (24/3)² = 88 − 64 = 24√24 = 4.89

x̄₁ = Σx/n₁ = 40/5 = 8, x̄₂ = Σx₂/n₂ = 24/3 = 8

7. AM, GM, HM — Relations
For Two Positive Observations
AM / GM = GM / HM ⟹ A/G = G/H ⟹ G² = A·H (by cross multiplication) ⟹ G =
±√(A·H)

i.e. the GM is itself the geometric mean of the AM and HM.

For More Than Two Observations (n observations)

AM ≥ GM ≥ HM
The equality sign holds true if and only if all the observations are equal.

Worked Example — Find HM given AM and GM

Given: AM = 25, GM = 15. Find HM.
A/G = G/H ⇒ 25/15 = 15/H ⇒ 25H = 15² = 225 ⇒ H = 225/25 = 9

Worked Example — Simple vs Weighted Harmonic Mean

Data: x = 1, 2, 4 with frequencies f = 4, 2, 1 respectively (N = Σf = 7).
Simple HM (unweighted, of the 3 distinct values 1, 2, 4):
H = 3 / (1 + 1/2 + 1/4) = 3 / (7/4) = 12/7 ≈ 1.71

Weighted HM (using the frequencies):

H = Σf / Σ(f/x) = 7 / (4/1 + 2/2 + 1/4) = 7 / 5.25 = 28/21 ≈ 1.33

Combined / Pooled / Composite Mean

For two groups of observations with sizes n₁, n₂ and means x̄₁, x̄₂:
x̄ (composite) = (n₁x̄₁ + n₂x̄₂) / (n₁ + n₂) = Σnᵢx̄ᵢ / Σnᵢ

Worked example: Group 1 (Darjeeling Tea): n₁ = 6, mean price P₁ = 600. Group 2 (Assam
Tea): n₂ = 10, mean price P₂ = 300.
P (composite) = (6×600 + 10×300) / (6+10) = (3600 + 3000)/16 = 6600/16 = 412.5
8. Correlation & Regression — Concepts
01/06/2026

Definition of Correlation
Correlation is the degree (strength) of association between two inter-dependent variables
(say x and y).

Functional Relationship
• y = f(x) → y is the dependent variable
• y = φ(x) → x is the independent variable
• Univariable: only one variable observed
• Paired-variate (bivariate) data: two variables observed together, e.g. (x, y) or (u, v)
• Multivariate data: two or more variables observed together, e.g. (x, y, z, u)

Scatter Plot (Scatter Diagram)

A scatter plot displays paired data (x, y) as points on a graph — x on the horizontal axis, y on
the vertical axis. The pattern of points indicates the type and strength of relationship.

• r₍ₓᵧ₎ > 0 — direct relationship: as x↑, y↑ (and as x↓, y↓). Points cluster around an
upward-sloping line.
• r₍ₓᵧ₎ < 0 — inverse relationship: as x↑, y↓ (and vice-versa). Points cluster around a
downward-sloping line.
• r₍ₓᵧ₎ = 0 — no linear relationship between the variables (points scattered with no clear
directional pattern).
9. Karl Pearson's Correlation Coefficient
Also called the product-moment correlation coefficient.
r₍ₓᵧ₎ = Cov(x, y) / (σₓ · σᵧ)

where Cov(x, y) = (1/n) Σ(xᵢ − x̄)(yᵢ − ȳ), and σₓ, σᵧ are the standard deviations of x and y.

Expanded (Computational) Formula

Cov(x, y) = Σxy/n − (Σx/n)(Σy/n)

σₓ = √[ Σx²/n − (Σx/n)² ] , σᵧ = √[ Σy²/n − (Σy/n)² ]

r₍ₓᵧ₎ = [ Σxy/n − (Σx/n)(Σy/n) ] / { √[Σx²/n − (Σx/n)²] · √[Σy²/n − (Σy/n)²] }

Multiplying numerator and denominator through by n² and simplifying gives the 'raw score'
formula used for direct computation:
r₍ₓᵧ₎ = [ nΣxy − (Σx)(Σy) ] / { √[nΣx² − (Σx)²] · √[nΣy² − (Σy)²] }

Computation Table Layout

x y xy x² y²

x₁ y₁ x₁y₁ x₁² y₁²

x₂ y₂ x₂y₂ x₂² y₂²

⋮ ⋮ ⋮ ⋮ ⋮

xₙ yₙ xₙyₙ xₙ² yₙ²

Σx Σy Σxy Σx² Σy²

10. Regression Coefficients
bᵧₓ = r · (σᵧ / σₓ)

bᵧₓ = Regression coefficient of y on x = r × (ratio of the SD of y to the SD of x).

This gives the relationship between the regression coefficient of y on x (bᵧₓ) and the
correlation coefficient (r): the regression coefficient of y on x is equal to the correlation
coefficient times the ratio of the standard deviation of y to that of x. The analogous
coefficient of x on y is bₓᵧ = r·(σₓ/σᵧ).
11. Properties (Theorems) of the Correlation Coefficient
Theorem 1
The correlation coefficient is independent of change of origin and scale of the observations.
i.e. if u = (x − A)/h and v = (y − B)/k (A, B = origin shift; h, k = scale factor), then:
r₍ₓᵧ₎ = r₍ᵤᵥ₎

Theorem 2
The correlation coefficient always lies between −1 and +1:
−1 ≤ r₍ₓᵧ₎ ≤ +1 (equivalently, |r₍ₓᵧ₎| ≤ 1)

Demonstration: for perfectly linearly related data, e.g. x = 1,...,10 and y = 10,...,1 (y = 11 −
x, a perfect inverse relation), the computation gives Σx = Σy = 55, Σx² = Σy² = 385, and the
correlation coefficient works out to exactly r = −1 — confirming the boundary of the range.
12. Worked Example — Correlation via Step-Deviation
(u, v)
04/06/2026

Transformed variates (change of origin and scale) are used to simplify computation when x,
y values are large:
u = (x − c) / (scale) [c = assumed origin], v = (y − c′) / (scale)

Computation Table Layout

x y xy x² y² u v u² v² uv

x₁ y₁ x₁y₁ x₁² y₁² u₁ v₁ u₁² v₁² u₁v₁

⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮

xₙ yₙ xₙyₙ xₙ² yₙ² uₙ vₙ uₙ² vₙ² uₙvₙ

Σu Σv Σu² Σv² Σuv

r₍ᵤᵥ₎ = Cov(u,v) / (σᵤ·σᵥ) = [Σuv/n − (Σu/n)(Σv/n)] / {√[Σu²/n − (Σu/n)²] · √[Σv²/n −

(Σv/n)²]}

Example Result
From one such worked table: r₍ᵤᵥ₎ = 0.74, obtained via the transformed variate computation
(equal, by Theorem 1, to the correlation coefficient of the original x, y data).

Another Worked Example

From a separate dataset, computation gave:
r₍ₓᵧ₎ = 0.84 (and, via a transformed u,v computation, r₍ᵤᵥ₎ = 0.92)

Interpretation: there is an 84% degree of association between x and y. The complementary

figure (1 − 0.84 = 0.16) represents about a 16% degree of dissociation between x and y.
13. Probability — Theorem of Compound Probability
Set notation used: ∩ = AND (intersection), ∪ = OR (union). A Venn diagram is used to
visualise events within a universal (super) set S.

Theorem 2 — Compound Probability

Statement: If two events A and B (defined on the same sample space) are independent, then
the probability of occurrence of the compound event (both A and B together) is given by:
P(A ∩ B) = P(A) · P(B) ⟹ A and B are independent

Otherwise, if A and B are not independent (i.e. dependent events):

P(A ∩ B) = P(A) · P(B|A) or P(A ∩ B) = P(B) · P(A|B)

Which form to use depends on which event's probability is already known — i.e. whether A is
treated as the predecessor (known first) or B is.
14. Spearman's Rank Correlation Coefficient
12/06/2026

Formula (No Tied Ranks)

ρ (or r₍ₓᵧ₎) = 1 − [ 6 Σd² / (n³ − n) ]

where d = R₁ − R₂ (difference between the two rankings of the same item), n = number of
items ranked.

Worked Problem — Musical Contest

Ten contestants in a musical contest were ranked by two judges A and B as follows.
Calculate Spearman's rank correlation coefficient.

Contestant 1 2 3 4 5 6 7 8 9 10

Judge A (R₁) 10 6 5 1 3 2 4 9 7 8

Judge B (R₂) 4 5 8 3 7 10 2 1 6 9

d = R₁ − R₂ −6 1 −3 −2 −4 −8 2 8 1 −1

d² 36 1 9 4 16 64 4 64 1 1

Σd² = 36+1+9+4+16+64+4+64+1+1 = 200

Solution
ρ = 1 − [6 × 200 / (10³ − 10)] = 1 − [1200/990] = 1 − 1.21 = −0.21

A rank correlation of −0.21 indicates a weak negative (inverse) association between the two
judges' rankings.

Formula with Tied Ranks (Correction Term)

R = 1 − [ 6{ Σd² − (1/12)Σ(t³ − t) } ] / (n³ − n)

where, for each group of tied ranks, t = number of items tied in that group; the correction
(1/12)Σ(t³−t) is summed over all tie-groups and subtracted from Σd² before applying the
main formula.

Worked Example with Ties (n = 10)

Given Σd² = 237, with tie groups of sizes 3, 2, 3, 2:
R = 1 − [6{237 − (1/12){(3³−3)+(2³−2)+(3³−3)+(2³−2)}}] / (10³−10)

= 1 − [6{237 − ½(24+6+24+6)}] / 990 = 1 − [6{237−30}] / 990

= 1 − [1422 − 30] / 990 = 1 − 1392/990 = 1 − 1.406 = −0.406 (approx.)

— End of Notes —