APPLIED STATISTICS

FORMULAS & SOLVED EXAMPLES

Department of Accounting & Information Systems — University of Barishal

📚 Topics Covered:
1. Measures of Central Tendency 5. Regression Analysis
2. Measures of Dispersion 6. Probability Theory
3. Skewness & Kurtosis 7. Sampling & Surveys
4. Correlation Analysis

CHAPTER 3: MEASURES OF CENTRAL TENDENCY

A. Formulas
ILO 3 — Definition, Types, and Uses of Central Tendency

[3.1] Population μ = ΣX / N Sum all values, divide by

Mean population size N

[3.2] Sample Mean X̄ = ΣX / n Sum all values, divide by sample

size n

[3.3] Median Position = (n+1) / 2 Find position of middle value in

Position ordered array

[3.4] Weighted Mean Xw = Σ(X·W) / ΣW W = weight assigned to each

value X

[3.5] Geometric GM = (X₁·X₂·...·Xn)^(1/n) Average percentage

Mean growth/change

[3.10] Grouped X̄ g = ΣfM / n f = frequency, M = midpoint of

Mean class

[3.11] Grouped Median = Lmd + [(n/2 - Lmd = lower boundary, F =

F) / fmd] × C cumulative freq before, fmd = freq
Median of median class, C = class width
 [3.12] Grouped Mode = Lmo + [Da / (Da + Lmo = lower boundary, Da = freq
Mode Db)] × C diff above, Db = freq diff below

B. Solved Math Examples

📊 Example 3.1 — Sample Mean (Ungrouped)

Data: Monthly sales ($000): 56, 67, 52, 45, 67 (n = 5)
Step 1: Add all values → 56 + 67 + 52 + 45 + 67 = 287
Step 2: X̄ = ΣX / n = 287 / 5 = 57.4
✅ Answer: The average monthly sales is $57,400

📊 Example 3.2 — Weighted Mean

Data: 2 gallons at $2.50, 0.5 gallon at $1.25, 1 gallon at $0.75, 3 gallons at $2.00
Step 1: Multiply quantity × price → 5.00 + 0.625 + 0.75 + 6.00 = 12.375
Step 2: Total weight (gallons) = 2 + 0.5 + 1 + 3 = 6.5
Step 3: Xw = Σ(X·W) / ΣW = 12.375 / 6.5 = $1.904 per gallon
✅ Answer: Weighted average cost = $1.90 per gallon

📊 Example 3.3 — Grouped Data Mean

Data: Ages of 50 CEOs — class: 50-54 (f=8), 55-59 (f=13), 60-64 (f=15), 65-69 (f=10), 70-74 (f=3),
75-79 (f=1)
Step 1: Find midpoints M: 52, 57, 62, 67, 72, 77
Step 2: Multiply f × M: 416 + 741 + 930 + 670 + 216 + 77 = 3050
Step 3: X̄ g = ΣfM / n = 3050 / 50 = 61.0
✅ Answer: Average CEO age = 61 years

CHAPTER 4: MEASURES OF DISPERSION

A. Formulas
ILO 4 — Dispersion, Coefficient of Range, Quartile Deviation, Standard Deviation, CV, Chebyshev

[3.6] Population σ² = Σ(Xi - μ)² / N Average of squared deviations

Variance from population mean
[3.7] Population Std σ = √σ² Square root of population variance
Dev

[3.8] Sample s² = Σ(Xi - X̄ )² / (n-1) Use n-1 (Bessel's correction) for

Variance sample data

[3.9] Sample Std s = √s² Square root of sample variance

Dev

[3.13] Grouped s² = [ΣfM² - nX̄ ²] / (n-1) f = freq, M = midpoint for grouped

Variance data

[3.15] Percentile Lp = (n+1) × P/100 P = desired percentile (e.g. 25 for

Location Q1)

[3.16] Chebyshev's At least (1 - 1/K²) × 100% K = number of std devs from

Theorem mean (K > 1)

[3.17] Coefficient of P = 3(X̄ - Median) / s Pearsonian: +ve = right skew, -ve

Skewness = left skew

[3.18] Coefficient of CV = (s / X̄ ) × 100 Relative variability — compare

Variation different datasets

B. Solved Math Examples

📊 Example 3.4 — Population Variance & Std Dev

Population data (weekly sales, $100s): 43.3, 54.2, 34.8, 42.9, 49.2, 29.5, 28.6
Step 1: μ = (43.3+54.2+34.8+42.9+49.2+29.5+28.6) / 7 = 282.5 / 7 = 40.36
Step 2: Squared deviations: (43.3-40.36)²+(54.2-40.36)²+...+(28.6-40.36)² = 461.14
Step 3: σ² = 461.14 / 7 = 65.88
Step 4: σ = √65.88 = 8.12
✅ Answer: Population variance = 65.88, Std Dev = 8.12

📊 Example 3.5 — Sample Variance & Std Dev

Sample data: 52, 48, 37, 54, 48, 15, 42, 12 (n = 8)
Step 1: X̄ = (52+48+37+54+48+15+42+12) / 8 = 308 / 8 = 38.5
Step 2: Σ(Xi - X̄ )² = (52-38.5)² + (48-38.5)² + ... + (12-38.5)² = 1775.0
 Step 3: s² = 1775.0 / (8-1) = 1775.0 / 7 = 253.57
Step 4: s = √253.57 = 15.93
✅ Answer: Sample variance = 253.57, Std Dev = 15.93

📊 Example 3.6 — Coefficient of Variation (Comparing Two Sets)

Dataset A: X̄ = 90, s = 9 | Dataset B: X̄ = 50, s = 8
Step 1: CVA = (9/90) × 100 = 10%
Step 2: CVB = (8/50) × 100 = 16%
✅ Answer: Dataset B has MORE relative variability (16% > 10%)
CV lets us compare variability across datasets with different means

📊 Example 3.7 — Chebyshev's Theorem

Mortgage rates: X̄ = 7.0%, s = 0.3%. At least what % fall within 2 std devs?
Step 1: K = 2 (2 standard deviations from mean)
Step 2: 1 - 1/K² = 1 - 1/4 = 0.75 → 75%
Range: 7.0 ± 2(0.3) = 6.4% to 7.6%
✅ Answer: At least 75% of mortgage rates fall between 6.4% and 7.6%

CHAPTER 5: SKEWNESS & KURTOSIS

A. Formulas
ILO 5 — Moments, Skewness, Kurtosis — Relative Measures

Skewness Sk = 3(X̄ - Median) / s Sk > 0: Right/positive skew | Sk <

(Pearsonian) 0: Left/negative skew

Moment-based α₃ = μ₃ / σ³ μ₃ = 3rd central moment, σ = std

Skewness dev

Kurtosis (Excess) α₄ = μ₄ / σ⁴ α₄ = 3: Normal (Mesokurtic); >3:

Leptokurtic; <3: Platykurtic

B. Solved Math Examples

📊 Example 5.1 — Pearsonian Coefficient of Skewness

 Data: Mortgage rates 7.1%, 7.3%, 7.0%, 6.9%, 6.6%, 6.9%, 6.5%, 7.3%, 6.85% (n=9)
Step 1: X̄ = (sum of all rates) / 9 = 62.45 / 9 = 6.939%
Step 2: Arrange in order: 6.5, 6.6, 6.85, 6.9, 6.9, 7.0, 7.1, 7.3, 7.3
Step 3: Median = middle value = 6.9 (5th position)
Step 4: s = 0.279 (from variance calculation)
Step 5: Sk = 3(6.939 - 6.9) / 0.279 = 3(0.039) / 0.279 = +0.42
✅ Answer: Sk = +0.42 → Slightly right-skewed (positive skew)

📊 Example 5.2 — Interpreting Kurtosis (Conceptual)

Std Dev (Machine A) = 0.15, Std Dev (Machine B) = 0.08
Higher kurtosis means more values cluster tightly near the mean (peaked)
Lower kurtosis means values are more spread (flat distribution)
✅ If Machine B has higher kurtosis → more consistent output → better quality control
Normal distribution has kurtosis = 3 (benchmark)

📊 Example 5.3 — Identifying Skewness Type (Quick Check)

Rule: If Mean > Median > Mode → RIGHT skew (positive)
Rule: If Mean < Median < Mode → LEFT skew (negative)
Rule: If Mean = Median = Mode → Symmetric (normal)
Example: X̄ = 57.4, Median = 56, Mode = 67
✅ Mean is between Median and Mode → distribution is slightly right-skewed

CHAPTER 6: CORRELATION ANALYSIS

A. Formulas
ILO 6 — Karl Pearson's r, Rank Correlation, Coefficient of Determination

[11.19] Total Sum of SST = Σ(Yi - Ȳ)² Total variation in Y from its mean
Squares

[11.20] Regression SSR = Σ(Ŷi - Ȳ)² Variation EXPLAINED by

Sum of Squares regression model

[11.21] Error Sum of SSE = Σ(Yi - Ŷi)² Variation NOT explained

Squares (residuals)
[11.22] Correlation r = √(SSR / SST) r measures strength of linear
Coeff (r) relationship

[11.23] r r = SSxy / √(SSx × SSy) SSxy = cross-products, SSx &

(Computational) SSy = sum of squares

[11.24] Coeff of r² = SSR / SST % of change in Y explained by X

Determination

[11.25] r² r² = (SSxy)² / (SSx × SSy) Easier computation of r²

(Computational)

SSx Formula SSx = ΣX² - (ΣX)²/n Sum of squared deviations for X

SSy Formula SSy = ΣY² - (ΣY)²/n Sum of squared deviations for Y

SSxy Formula SSxy = ΣXY - (ΣX)(ΣY)/n Sum of cross-product deviations

B. Solved Math Examples

📊 Example 6.1 — Pearson Correlation (Hop Scotch Airlines)

Data: n=15 months, Advertising (X) vs Passengers (Y)
Given: ΣX=187, ΣY=268, ΣX²=2469, ΣY²=4960, ΣXY=3490
Step 1: SSx = 2469 - (187)²/15 = 2469 - 2331.27 = 137.73
Step 2: SSy = 4960 - (268)²/15 = 4960 - 4788.27 = 171.73
Step 3: SSxy = 3490 - (187×268)/15 = 3490 - 3341.07 = 148.93
Step 4: r = 148.93 / √(137.73 × 171.73) = 148.93 / 153.82 = 0.9683
✅ Answer: r = 0.97 → STRONG POSITIVE relationship between advertising & passengers

📊 Example 6.2 — Coefficient of Determination r²

Continuing from Example 6.1 (r = 0.9683)
Method 1: r² = (0.9683)² = 0.9376 → 93.8%
Method 2: r² = (SSxy)² / (SSx × SSy) = (148.93)² / (137.73 × 171.73)
= 22180.07 / 23659.38 = 0.9376
✅ Answer: 93.8% of the change in passengers is explained by advertising spending
Only 6.2% is unexplained (due to other factors)
 📊 Example 6.3 — Interpreting r Value
r = +1.00 → Perfect positive relationship (all points on a straight line going up)
r = +0.90 → Strong positive relationship
r = +0.50 → Moderate positive relationship
r ≈ 0.00 → No linear relationship
r = -0.80 → Strong negative relationship
r = -1.00 → Perfect negative relationship (all points on a straight line going down)
✅ Rule: |r| > 0.7 → Strong | 0.4-0.7 → Moderate | < 0.4 → Weak

CHAPTER 7: REGRESSION ANALYSIS

A. Formulas
ILO 7 — Simple Linear Regression, Least Squares, Regression Line, Coefficient

Regression Model Ŷ = b₀ + b₁X b₀ = Y-intercept, b₁ = slope

(regression coefficient)

[11.7] Error Term e = Yi - Ŷi Difference between actual and

predicted Y

[11.8] OLS Criterion Σ(Yi - Ŷi)² = minimum OLS finds the line that minimizes
sum of squared errors

[11.9] SSx SSx = ΣX² - (ΣX)²/n Sum of squares for X (same as in

correlation)

[11.10] SSy SSy = ΣY² - (ΣY)²/n Sum of squares for Y

[11.11] SSxy SSxy = ΣXY - (ΣX)(ΣY)/n Cross-product sum of squares

[11.12] Slope (b₁) b₁ = SSxy / SSx Regression coefficient — change

in Y per unit of X

[11.13] Intercept (b₀) b₀ = Ȳ - b₁X̄ Y-intercept of regression line

B. Solved Math Examples

📊 Example 7.1 — Full OLS Regression (Hop Scotch Airlines)

Data: n=15, ΣX=187, ΣY=268, ΣX²=2469, ΣY²=4960, ΣXY=3490
Step 1: SSx = 2469 - (187)²/15 = 137.73
Step 2: SSxy = 3490 - (187×268)/15 = 148.93
Step 3: b₁ = SSxy / SSx = 148.93 / 137.73 = 1.08
Step 4: Ȳ = 268/15 = 17.87 | X̄ = 187/15 = 12.47
Step 5: b₀ = Ȳ - b₁X̄ = 17.87 - 1.08(12.47) = 17.87 - 13.47 = 4.40
Step 6: Regression Model: Ŷ = 4.40 + 1.08X
✅ Meaning: If advertising = $10,000 → Passengers = 4.40 + 1.08(10) = 15.2 thousand

📊 Example 7.2 — Predicting with Regression

Regression model: Ŷ = 4.40 + 1.08X (X = advertising in $000, Y = passengers in 000)
Predict passengers if advertising = $11,000:
Ŷ = 4.40 + 1.08(11) = 4.40 + 11.88 = 16.28 thousand
Predict passengers if advertising = $15,000:
Ŷ = 4.40 + 1.08(15) = 4.40 + 16.20 = 20.60 thousand
✅ Interpretation: Each additional $1,000 in advertising brings ~1,080 more passengers

📊 Example 7.3 — GPA vs Job Offers (Practice)

Data: n=10 students, ΣGPA=26.96, ΣOffers=28, ΣGPAxOffers=80.78, ΣGPA²=75.97
Step 1: SSx = 75.97 - (26.96)²/10 = 75.97 - 72.68 = 3.29
Step 2: SSxy = 80.78 - (26.96×28)/10 = 80.78 - 75.49 = 5.29
Step 3: b₁ = 5.29 / 3.29 = 1.61 offers per GPA point
Step 4: b₀ = (28/10) - 1.61(26.96/10) = 2.8 - 1.61(2.696) = -1.54
Step 5: Ŷ = -1.54 + 1.61(GPA)
✅ If GPA = 3.22: Ŷ = -1.54 + 1.61(3.22) ≈ 3.6 → predict ~4 job offers

CHAPTER 8: PROBABILITY THEORY

A. Formulas
ILO 8 — Classical, Relative Frequency, Rules of Probability, Bayes' Theorem

Classical Probability P(A) = f / N f = favorable outcomes, N = total

equally likely outcomes
 P(A) = frequency / total Based on observed data
Relative Frequency (empirical probability)
trials

Complement Rule P(A') = 1 - P(A) Probability of event NOT occurring

Conditional P(A|B) = P(A ∩ B) / P(B) Probability of A given B has

Probability occurred

[4.4] Multiplication P(A ∩ B) = P(A) × P(B) For INDEPENDENT events — no

(Independent) influence on each other

[4.5] Multiplication P(A ∩ B) = P(A) × P(B|A) For DEPENDENT events — use

(Dependent) conditional probability

[4.6] Addition (Not P(A ∪ B) = P(A) + P(B) - Subtract overlap to avoid double
Mutually Exclusive) P(A ∩ B) counting

[4.7] Addition P(A ∪ B) = P(A) + P(B) Events cannot happen at same

(Mutually Exclusive) time — no overlap

P(A|B) = P(A)·P(B|A) / Update probability based on new

Bayes' Theorem information
ΣP(Ai)·P(B|Ai)

B. Solved Math Examples

📊 Example 8.1 — Multiplication Rule (Independent Events)

Problem: P(car stops at motel) = 0.20. What is P(next 2 cars both stop)?
Events are independent (each car's decision doesn't affect the other)
Step 1: P(S₁) = 0.20, P(S₂) = 0.20
Step 2: P(S₁ ∩ S₂) = P(S₁) × P(S₂) = 0.20 × 0.20 = 0.04
✅ Answer: 4% probability that both cars will stop

📊 Example 8.2 — Addition Rule (Not Mutually Exclusive)

Problem: Gasoline preferences — P(Regular)=0.20, P(Ethanol)=0.20, P(Regular AND Ethanol)=0.05
Events are NOT mutually exclusive (a customer can prefer Regular with Ethanol)
Step 1: P(R ∪ E) = P(R) + P(E) - P(R ∩ E)
 Step 2: P(R ∪ E) = 0.20 + 0.20 - 0.05 = 0.35
✅ Answer: 35% probability the customer prefers Regular OR Ethanol

📊 Example 8.3 — Bayes' Theorem

Machine A makes 60% of output (2% defective); Machine B makes 40% (4% defective)
A defective unit is found. What is P(it came from Machine A)?
Step 1: P(A) = 0.60, P(D|A) = 0.02 → P(A ∩ D) = 0.60 × 0.02 = 0.012
Step 2: P(B) = 0.40, P(D|B) = 0.04 → P(B ∩ D) = 0.40 × 0.04 = 0.016
Step 3: P(D) = 0.012 + 0.016 = 0.028
Step 4: P(A|D) = P(A ∩ D) / P(D) = 0.012 / 0.028 = 0.4286
✅ Answer: 42.9% chance the defective unit came from Machine A

CHAPTER 9: SAMPLING & SAMPLING DISTRIBUTIONS

A. Formulas
ILO 9 — Standard Error, Central Limit Theorem, Z-scores for Sampling

[6.1] Mean of X̄̄ = ΣX̄ / K K = number of samples drawn

Sample Means

[6.3] Standard Error σx̄ = √σ²x̄ Std dev of sampling distribution of

(σx̄ ) means

[6.4] Standard Error σx̄ = σ / √n σ = population std dev, n = sample

Formula size

[6.5] Finite σx̄ = (σ/√n) × √[(N-n)/(N- Use when n > 5% of population N

Population 1)]
Correction

[6.6] Z-score for Z = (X̄ - μ) / σx̄ Standard normal deviate for

Means sample means

[6.8] Std Error of σp = √[p(1-p)/n] p = population proportion

Proportions

[6.10] Z-score for Z = (p̂ - p) / σp p̂ = sample proportion, p =

 Proportions population proportion

B. Solved Math Examples

📊 Example 9.1 — Standard Error and Z-score

Population: mean returns = 30%, std dev = 12%. Sample of n = 250 firms.
Find P(sample mean > 31%)
Step 1: Standard Error = σ/√n = 12/√250 = 12/15.81 = 0.759%
Step 2: Z = (X̄ - μ) / σx̄ = (31 - 30) / 0.759 = 1.32
Step 3: P(Z > 1.32) = 0.5000 - 0.4066 = 0.0934
✅ Answer: 9.34% chance that sample mean exceeds 31%

📊 Example 9.2 — Proportion Z-score

22% of firms market directly to consumers. Sample n = 250 firms.
Find P(sample proportion > 20%)
Step 1: σp = √[p(1-p)/n] = √[(0.22)(0.78)/250] = √0.000686 = 0.0262
Step 2: Z = (p̂ - p) / σp = (0.20 - 0.22) / 0.0262 = -0.76
Step 3: P(p̂ > 0.20) = 0.5000 + 0.2764 = 0.7764
✅ Answer: 77.64% chance that more than 20% of sample firms do direct marketing

📊 Example 9.3 — Central Limit Theorem Application

Mean deposits at a bank: μ = $7,012, σ = $532. Sample n = 35 deposits.
Find P(sample mean > $6,911)
Step 1: σx̄ = σ/√n = 532/√35 = 532/5.916 = 89.93
Step 2: Z = (X̄ - μ) / σx̄ = (6911 - 7012) / 89.93 = -101/89.93 = -1.12
Step 3: P(X̄ > 6911) = P(Z > -1.12) = 0.5 + P(0 < Z < 1.12) = 0.5 + 0.3686 = 0.8686
✅ Answer: 86.86% probability that sample mean exceeds $6,911

QUICK FORMULA REFERENCE SUMMARY

All key formulas from the syllabus at a glance:

Topic Formula Meaning Chapter

Population Mean μ = ΣX / N Average of all population Ch. 3
values
Sample Mean X̄ = ΣX / n Average of sample values Ch. 3
Population σ² = Σ(X-μ)²/N Spread around population Ch. 4
Variance mean
Sample s² = Σ(X-X̄ )²/(n-1) Spread around sample mean Ch. 4
Variance
Std Deviation σ or s = √variance Square root of variance Ch. 4
Coeff. of CV = (s/X̄ ) × 100 Relative variability % Ch. 4/5
Variation
Coeff. of Sk = 3(X̄ -Med)/s Symmetry of distribution Ch. 5
Skewness
Pearson's r r = SSxy/√(SSx·SSy) Linear correlation strength Ch. 6
Coeff. of r² = (SSxy)²/(SSx·SSy) % variation explained Ch. 6
Determination
Regression b₁ = SSxy / SSx Change in Y per unit X Ch. 7
Slope
Regression b₀ = Ȳ - b₁X̄ Y-value when X = 0 Ch. 7
Intercept
P (Independent) P(A∩B) = P(A)×P(B) Joint prob — independent Ch. 8
P (Dependent) P(A∩B) = P(A)×P(B|A) Joint prob — dependent Ch. 8
P (Addition) P(A∪B) = P(A)+P(B)-P(A∩B) Either A or B Ch. 8
Standard Error σx̄ = σ / √n Std dev of sample means Ch. 9
Z-score (Means) Z = (X̄ - μ) / σx̄ Normal deviate for means Ch. 9

Applied Statistics — Formula Booklet | University of Barishal