Business Analytics

Unit 2 — Statistics & Sampling

Complete Study Notes with Solved Numericals

Topics: Population & Sample • Descriptive Statistics • Variance & SD Frequency

Distributions • Sampling Methods • Confidence Intervals Standard Error • Regression •
Hypothesis Testing • Data Visualization

MCA Exam Preparation

 Table of Contents

Unit 2 — All Topics at a Glance

1 Population, Sample & Estimators

2 Descriptive Statistics — Mean, Median, Mode, Dispersion

3 Variance & Standard Deviation — Theory + 6 Solved Numericals

4 Frequency Distributions

5 Sampling Methods

6 Standard Error of the Mean

7 Confidence Intervals

8 Regression Models

9 Hypothesis Testing

10 Data Visualization

11 Quick Formulas Reference Sheet

 1. Population, Sample & Estimators

The foundation — understanding where data comes from

Core Definitions
Term Definition Example

Population ALL items of interest for a decision All 10,000 students in a university

Sample A SUBSET of the population 200 randomly selected students

Census Data from the ENTIRE population Surveying all 10,000 students

Most business data is sample data — collecting population data is too expensive, too slow, or impossible.

Types of Estimators
Point Estimator: A single number from sample data that estimates a population value. E.g., sample
mean estimates population mean.
Unbiased Estimator: An estimator whose expected value equals the true population parameter — it
doesn't systematically over or under-estimate.
Interval Estimate: A range of values believed to contain the true population parameter. More realistic
than a single number.
Confidence Interval: Interval estimate that comes with a probability. E.g., '95% confident the true mean
lies between X and Y.'

Exam Tip: Population = All. Sample = Subset. Census = Population data. Businesses mostly use samples.
Point estimator = single number. Interval estimate = a range.
 2. Descriptive Statistics

Numbers that summarise and describe your dataset

Measures of Central Location

Measure What it is Formula Best when

Mean (Arithmetic) The classic average Sum of values / N Data has no extreme outliers

Median Middle value when sorted Middle value (or avg of 2 middles) Data has outliers

Mode Most frequently occurring value Value with highest frequency Categorical data

Measures of Dispersion (Spread)

Dispersion: Degree of variation/spread in data. Two datasets with same mean can have very different
spreads.
Standard Deviation (σ): Most common measure of spread. Same unit as data. Small SD = tight data.
Large SD = spread out data.
Variance (σ²): Standard deviation squared. Used in calculations but unit is 'squared'. σ² = Σ(x−µ)² / N
Coefficient of Variation (CV): Relative dispersion = (SD / Mean) × 100. Used to compare spread across
different datasets or units.
Coefficient of Skewness (CS): Measures asymmetry of data around the mean. Zero = symmetric.
Positive = right tail. Negative = left tail.
z-score: Standardized value = (x − µ) / σ. Tells how many SDs a value is from the mean.

The Empirical (Sigma) Rule — for normal distributions

Range % of Data Covered Example (Mean=50, SD=10)

Mean ± 1σ ~68% 40 to 60

Mean ± 2σ ~95% ← exam favourite 30 to 70

Mean ± 3σ ~99.7% 20 to 80

Exam Tip: 95% = Two sigma limit. Mean is the arithmetic mean. Median = middle. Mode = most frequent. CV =
relative dispersion. CS = skewness. z-score = standardized value.
 3. Variance & Standard Deviation

Theory, step-by-step method, and 6 fully solved numericals

What Do They Measure?

Both measure how spread out data is from its mean. Two datasets can have the same average but
completely different spreads — Variance and SD capture that difference.

Example: Student A scored 48,50,52,50,50 and Student B scored 10,90,20,80,50. Both have mean = 50. But A
is consistent, B is all over the place. Variance/SD measures this difference.

The Formulas

Variance: sigma^2 = SIGMA(x - mu)^2 / N

Standard Deviation: sigma = SQRT[ SIGMA(x - mu)^2 / N ]

Where: x = each value

mu = mean of all values
N = total number of values
SIGMA = sum of all

The 6-Step Method — Always Follow This Order

Find the Mean (µ)
1
Add all values and divide by N

Find each Deviation

2
Subtract mean from each value: (x − µ)

Square each Deviation

3
Multiply deviation by itself: (x − µ)² — removes negatives

Sum all Squared Deviations

4
Add all squared deviations: Σ(x − µ)²

Divide by N → Variance
5
σ² = Σ(x − µ)² / N

Take Square Root → SD

6
σ = √σ² — brings unit back to original

Exam Tip: Always verify Step 2: deviations must sum to exactly 0. If not, your mean is wrong — go back and
recalculate!

Why square the deviations?

Because deviations always sum to 0 (positives and negatives cancel). Squaring makes all values positive
and amplifies larger deviations — a value 10 units away contributes 100 to the sum, while a value 2 units
away contributes only 4.
Numerical 1 — Foundation Level
Dataset: 3, 5, 7, 9, 11 | N = 5 | Find Mean, Variance, and Standard Deviation

Step 1 — Mean

Sum = 3 + 5 + 7 + 9 + 11 = 35
N = 5
mu = 35 / 5 = 7

Steps 2, 3, 4 — Deviation Table

Value (x) Deviation (x - 7) Squared (x - 7)²

3 3 - 7 = -4 (-4)² = 16

5 5 - 7 = -2 (-2)² = 4

7 7 - 7 = 0 0² = 0

9 9 - 7 = +2 (+2)² = 4

11 11 - 7 = +4 (+4)² = 16

Sum = 35 Sum = 0 ✓ SIGMA = 40

Step 5 — Variance

sigma² = 40 / 5 = 4

Step 6 — Standard Deviation

sigma = SQRT(4) = 2

FINAL ANSWERS:
Mean (mu) = 7
Variance (sigma²) = 4
Std Deviation (sigma) = 2

Meaning: On average, each value is 2 units away from the mean of 7.

Numerical 2 — Exam Style (Student Marks)
Marks of 6 students: 12, 18, 24, 30, 36, 42 | N = 6 | Find variance and SD

Step 1 — Mean

Sum = 12 + 18 + 24 + 30 + 36 + 42 = 162
N = 6
mu = 162 / 6 = 27

Shortcut: For arithmetic sequence, mean = (first + last)/2 = (12+42)/2 = 27

checkmark

Steps 2, 3, 4 — Deviation Table

Value (x) Deviation (x - 27) Squared (x - 27)²

12 12 - 27 = -15 (-15)² = 225

18 18 - 27 = -9 (-9)² = 81

24 24 - 27 = -3 (-3)² = 9

30 30 - 27 = +3 (+3)² = 9

36 36 - 27 = +9 (+9)² = 81

42 42 - 27 = +15 (+15)² = 225

Sum = 162 Sum = 0 ✓ SIGMA = 630

Smart shortcut: Squared values are symmetric — 225, 81, 9, 9, 81, 225. Calculate only 3 unique values: Σ = 2 ×
(225+81+9) = 2 × 315 = 630.

Variance: sigma² = 630 / 6 = 105

Std Dev: sigma = SQRT(105)
= SQRT(4 x 26.25) ... not clean
sigma = approx 10.25

Quick check: 10^2 = 100, 10.5^2 = 110.25 ... so SQRT(105) is between = 10.25
checkmark

FINAL ANSWERS:
Mean (mu) = 27 marks
Variance (sigma²) = 105
Std Deviation (sigma) = 10.25 marks
Numerical 3 — Heights (Clean SD)
Heights of 7 students (cm): 150, 155, 160, 165, 170, 175, 180 | N = 7

Step 1 — Mean
Sum = 150+155+160+165+170+175+180 = 1155
mu = 1155 / 7 = 165

Shortcut: 7 evenly spaced values -> middle (4th) value = mean = 165 checkmark

Deviation Table

Height (x) Deviation (x - 165) Squared (x - 165)²

150 -15 225

155 -10 100

160 -5 25

165 0 0

170 +5 25

175 +10 100

180 +15 225

Sum = 1155 Sum = 0 ✓ SIGMA = 700

Symmetry shortcut: SIGMA = 2 × (225+100+25) + 0 = 2 × 350 = 700. Only calculate half the table!

Variance: sigma² = 700 / 7 = 100

Std Dev: sigma = SQRT(100) = 10 (clean whole number!)

FINAL ANSWERS:
Mean (mu) = 165 cm
Variance (sigma²) = 100 cm²
Std Deviation (sigma) = 10 cm

Verification: mu ± 2*sigma = 145 to 185. All 7 values fall in this range.

checkmark
Numerical 4 — Comparing Two Datasets
Two batsmen's scores over 5 matches: Batsman A: 40, 50, 60, 70, 80 Batsman B: 20, 40, 60, 80, 100 Who is
more consistent?

This is the most common exam type — same mean, find who is more consistent using SD.

Step 1 — Mean for both:

Batsman A: (40+50+60+70+80) / 5 = 300 / 5 = 60
Batsman B: (20+40+60+80+100) / 5 = 300 / 5 = 60

Both average 60 runs. SD will tell us who is more consistent.

Batsman A — Deviation Table (mean = 60)

Score (x) Deviation (x - 60) Squared (x - 60)²

40 -20 400

50 -10 100

60 0 0

70 +10 100

80 +20 400

Sum=300 0✓ SIGMA = 1000

sigma²_A = 1000 / 5 = 200

sigma_A = SQRT(200) = SQRT(100 x 2) = 10*SQRT(2) = 14.14

Batsman B — Deviation Table (mean = 60)

Score (x) Deviation (x - 60) Squared (x - 60)²

20 -40 1600

40 -20 400

60 0 0

80 +20 400

100 +40 1600

Sum=300 0✓ SIGMA = 4000

sigma²_B = 4000 / 5 = 800

sigma_B = SQRT(800) = SQRT(400 x 2) = 20*SQRT(2) = 28.28

Batsman A Batsman B

Mean 60 60

Variance 200 800

Std Dev 14.14 28.28

 Consistency MORE consistent (lower SD) Less consistent (higher SD)

CONCLUSION: Batsman A is MORE CONSISTENT

Batsman A: sigma = 14.14 (scores stay close to average)
Batsman B: sigma = 28.28 (scores swing wildly 20 to 100)

Lower SD = more reliable performance. In business, lower SD = lower risk.

Numerical 5 — Frequency Distribution Table
Score (x): 10, 20, 30, 40, 50 | Frequency (f): 3, 5, 8, 3, 1 Find variance and standard deviation using the
frequency formula.

Modified Formula for Frequency Data

Mean: mu = SIGMA(f * x) / SIGMA(f)

Variance: sigma² = SIGMA[ f * (x - mu)² ] / SIGMA(f)
Std Dev: sigma = SQRT(sigma²)

KEY DIFFERENCE: Multiply squared deviations by frequency (f) before summing.

Step 1 — Find Mean using f×x

Score (x) Frequency (f) f×x

10 3 30

20 5 100

30 8 240

40 3 120

50 1 50

Total SIGMA f = 20 SIGMA fx = 540

mu = 540 / 20 = 27

Steps 2-4 — Full Weighted Deviation Table

x f (x-27) (x-27)² f × (x-27)²

10 3 -17 289 867

20 5 -7 49 245

30 8 +3 9 72

40 3 +13 169 507

50 1 +23 529 529

— 20 ~0 ✓ — SIGMA = 2220

Variance: sigma² = 2220 / 20 = 111

Std Dev: sigma = SQRT(111) = approx 10.54

Check: 10^2=100, 11^2=121 -> SQRT(111) is between 10 and 11, closer to 10.5
checkmark
FINAL ANSWERS:
Total observations (SIGMA f) = 20
Mean (mu) = 27
Variance (sigma²) = 111
Std Deviation (sigma) = 10.54
Numerical 6 — Salaries with Decimal Mean
Monthly salaries (Rs. 000s): 25, 30, 35, 40, 45, 50 | N = 6 This tests handling of decimal mean values carefully.

Step 1 — Mean:
Sum = 25+30+35+40+45+50 = 225
mu = 225 / 6 = 37.5

Deviation Table (careful with decimals!)

Salary (x) Deviation (x - 37.5) Squared (x - 37.5)²

25 -12.5 156.25

30 -7.5 56.25

35 -2.5 6.25

40 +2.5 6.25

45 +7.5 56.25

50 +12.5 156.25

Sum=225 0✓ SIGMA = 437.50

WATCH OUT: (-7.5)² = 56.25, NOT 52.25. A common mistake with decimal squaring. Always double-check your
arithmetic here.

Variance: sigma² = 437.50 / 6 = 72.92

Std Dev: sigma = SQRT(72.92) = 8.54 (Rs. 000s) = Rs. 8,540

FINAL ANSWERS:
Mean (mu) = 37.5 (Rs. 37,500)
Variance (sigma²) = 72.92
Std Dev (sigma) = 8.54 (Rs. 8,540)

Meaning: Salaries typically differ by Rs.8,540 from the average of Rs.37,500.

 4. Frequency Distributions

Organising raw data into groups to see patterns

Key Terms
Class/Interval: Each group or bucket (e.g. 0-9, 10-19). Classes must not overlap.
Class Width: Gap between lower limits of consecutive classes. E.g. 10-0 = 10 (NOT 9, NOT 39).
Relative Frequency: Proportion of total in a class = Class count / Total count. All must sum to 1.0.
Cumulative Frequency: Running total of frequencies as you move down the table.
Midpoint: (Lower bound + Upper bound) / 2. E.g. midpoint of 30-39 = (30+39)/2 = 34.5

Worked Example — Hours worked by 400 students

Interval Count Relative Freq (÷400) Cumulative Freq Midpoint

0-9 20 0.05 20 4.5

10-19 80 0.20 100 14.5

20-29 200 0.50 300 24.5

30-39 100 0.25 400 34.5

• Students working 19 hours or less = 20 + 80 = 100

• Relative frequency for 9 hrs or less = 20 / 400 = 0.05
• Percentage working 19 hrs or less = 100/400 = 0.25 (25%)
• Class width = 10 (gap between lower bounds: 10-0=10, NOT 9-0=9)

Exam Tip: Class width = gap between consecutive lower limits. Midpoint of 30-39 = 34.5 (not 35). Relative
frequencies always add up to exactly 1.0.
 5. Sampling Methods

Different ways to pick a representative sample

Overview of All 5 Methods

Method How it Works Key Advantage Exam Keyword

Simple Random
Every item has equal chance. Use random number table.
Easy, unbiased Random table

Systematic Pick every nth item from list. Fast, structured Every nth item

Stratified Divide into subgroups (strata), sample from [Link] representation Subgroups/strata

Cluster
Divide into natural clusters, select whole clusters randomly.
Cost-effective, geographic Natural subsets

Judgment Expert manually selects sample. Subjective.

Quick when expertise exists Expert picks

Important Points
• Most common probabilistic method = Simple Random Sampling
• Subjective method = Judgment Sampling (only one that is NOT probabilistic)
• Sampling Plan: A formal description of how samples will be obtained — created BEFORE data
collection begins
• Pros of Simple Random: Easy to implement, unbiased
• Pros of Stratified: Represents all subgroups evenly across all strata
• Cluster vs Stratified: Cluster = select whole groups. Stratified = sample from within each group.

Exam Tip: Cluster = geographic natural subsets (select whole clusters). Stratified = equal representation
(sample from within each stratum). Simple random = random number table. Systematic = every nth item.
 6. Standard Error of the Mean

How reliable is your sample mean?

Concept
Standard Error (SE) measures how much sample means are expected to vary from the true population
mean. A small SE means your sample mean is very close to the true population mean — it is reliable.

Analogy: If you shoot arrows at a target, SE measures how tightly clustered the arrows are. Small SE = arrows
cluster tightly (consistent). Large SE = arrows spread widely (unreliable).

Formula

SE = Population SD / SQRT(n)

Where: Population SD = standard deviation of the population

n = sample size

As n increases -> SQRT(n) increases -> SE decreases -> estimate improves

Exam Calculations (Population SD = 2.89, Variance = 8.33)

Sample size (n) SQRT(n) SE = 2.89 / SQRT(n) Reliability

10 3.162 0.914 Least reliable

20 4.472 0.646 Getting better

25 5.000 0.577 Answer = 0.577

100 10.00 0.289 Very reliable

500 22.36 0.129 Most reliable

Key insight: To halve the SE, you need to quadruple the sample size (not just double it). Sampling has
diminishing returns.

Exam Tip: SE = Population SD / SQRT(n). Larger sample = smaller SE = more accurate. To halve SE,
quadruple n. n=25 -> SE=0.577 (exam answer = option 4).
 7. Confidence Intervals

A realistic range for where the true population value lies

What is a Confidence Interval?

Instead of giving a single estimate, a Confidence Interval (CI) gives a range that accounts for uncertainty.
E.g., '95% confident the true average lies between Rs.10,777 and Rs.14,483.'

What does '95% confident' mean? If you repeated the sampling process 100 times, approximately 95 of
those intervals would contain the true mean. It does NOT mean 95% probability that the true mean is in THIS
specific interval.

When to Use Which Distribution

Situation Use

Population SD is KNOWN z-distribution (normal curve)

Population SD is UNKNOWN t-distribution ← used in your exam

Formula (Unknown Population SD — t-distribution)

CI = Mean +- t * (SD / SQRT(n))

Upper Limit = Mean + t * (SD / SQRT(n))

Lower Limit = Mean - t * (SD / SQRT(n))

Exam values: Mean = 12,630 | SD = 5,393

Exam Answers — Memorise These

n Upper Limit Lower Limit Width Note

20 15,154 10,106 5,048 Widest — least precise

25 14,856 10,404 4,452

30 14,644 10,616 4,028

35 14,483 10,777 3,706

50 14,163 11,097 3,066 Narrowest — most precise

Key Pattern: As n increases from 20 to 50, interval gets narrower. Larger sample = more precision =
narrower CI.

Exam Tip: Unknown SD -> t-distribution. Larger n = narrower CI = more precise. Level of significance (alpha) =
risk of being wrong. For 95% CI, alpha = 0.05.
 8. Regression Models

Predicting one variable from others using a math equation

Concept
Dependent variable: The outcome you are predicting (e.g., bank balance, sales)
Independent variables: The inputs that influence the outcome (e.g., age, education, wealth)
Regression: Finds the mathematical relationship between dependent and independent variables

Bank Balance Model (from your exam)

Balance = -17,732 + 367*Age + 1,300*Education + 0.116*Wealth

Reading the equation:

-17,732 = base value (when all inputs = 0)
+367 per year of age -> older = higher balance
+1,300 per year of edu -> more educated = higher balance
+0.116 per rupee of wealth -> wealthier = higher balance

Solved Calculations
Age Education Wealth Calculation Balance

32 16 -17732
1,50,000
+ (367x32) + (1300x16) + (0.116x150000) = -17732+11744+20800+17400
32,212

32 16 1,00,000 -17732 + 11744 + 20800 + (0.116x100000=11600) 26,412

30 16 1,50,000 -17732 + (367x30=11010) + 20800 + 17400 31,478

32 16 50,000 -17732 + 11744 + 20800 + (0.116x50000=5800) 20,612

52 16 1,50,000 -17732 + (367x52=19084) + 20800 + 17400 39,552

Exam Tip: Use 0.116 for wealth (NOT 116!). Calculate each term separately before adding. Education x 1300
is the biggest contributor here.
 9. Hypothesis Testing

Using data to test whether a claim about a population is true

The Two Hypotheses

Null Hypothesis (H0): The current/accepted belief. Assumed true unless strong evidence proves
otherwise. Usually states 'no effect' or 'no difference.'
Alternative Hypothesis (H1): What you are trying to prove. The complement of H0 — if H0 is false, H1
must be true.

Two Types of Errors — Must Know

Error Type What Happened Real-world Meaning Probability

Type I Error H0 is TRUE but you REJECTED

Falseitalarm — you think drug works but it doesn't
alpha (level of significance)

Type II Error H0 is FALSE but you KEPT

Missed
it it — drug works but you concluded it doesn't beta

Level of Significance (alpha): Acceptable risk of Type I error. Typically 0.05 (5%). Defined BEFORE the
test.
Power of the Test: 1 - beta. Probability of correctly detecting that H0 is false. Higher = better.
Confidence Level: 1 - alpha. For alpha=0.05, confidence level = 95%.

The 5-Step Procedure — In Exact Order

Step Action What it Means

1 Identify parameter, formulate H0 and H1 What are you testing? State both hypotheses clearly.

2 Select level of significance (alpha) How much risk of error can you tolerate? Usually 0.05.

3 Determine decision rule Define when you will reject H0.

4 Collect data, calculate test statistic Run the experiment, compute the statistic.

5 Apply decision rule, draw conclusion Reject or fail to reject H0. State conclusion clearly.

Which Test to Use?

Situation Test

Variance KNOWN z-test (two-sample test for means)

Variance UNKNOWN/unequal t-test (two-sample test for means)

Market share, proportions Proportion test

Exam Tip: Type I = rejecting a TRUE H0 (false alarm). Type II = keeping a FALSE H0 (missed). Power =
1-beta. Step order is FIXED — memorise it. z-test = known variance. t-test = unknown variance.
 10. Data Visualization

Turning raw numbers into visual insights

Key Definitions
Data Visualization: Process of displaying data in a meaningful visual fashion to provide insights that
support better decisions.
Dashboard: A single visual display of multiple key business measures (KPIs). Like a car dashboard —
one screen, all vital information.
Descriptive Statistics: Methods of describing and summarising data using tabular, visual, and
quantitative techniques.
Frequency Distribution: A table showing the number of observations in each of several non-overlapping
groups.
Most Used Software: Tableau (most widely used data visualization tool).

Chart Types — When to Use Each

Chart Type Best Used For Key Insight

Line Chart Data over time Trends — going up, down, or fluctuating?

Pie Chart Parts of a whole Proportion/share of each category

Bar/Column Comparing categories Which category is biggest/smallest?

Scatter Chart Two numeric variables Relationship between X and Y?

Area Chart Trend + volume Combines features of pie + line chart

Histogram Distribution of continuous data How is data spread across intervals?

Excel Naming — Important for exam: In Excel, vertical bar charts are called Column charts and horizontal
bar charts are called Bar charts.

Exam Tip: Dashboard = visual display of key measures. Tableau = top tool. Pie = proportions. Scatter =
2-variable relationship. Line = time trends. Excel: vertical = Column chart, horizontal = Bar chart.
 11. Quick Formulas Reference Sheet

Everything you need in one place — review before the exam

All Key Formulas

DESCRIPTIVE STATISTICS
Mean: mu = SIGMA(x) / N
Variance: sigma^2 = SIGMA(x - mu)^2 / N
Std Deviation: sigma = SQRT[ SIGMA(x - mu)^2 / N ]
z-score: z = (x - mu) / sigma
CV: CV = (sigma / mu) x 100

FREQUENCY TABLE (weighted)

Mean: mu = SIGMA(f*x) / SIGMA(f)
Variance: sigma^2 = SIGMA[ f*(x-mu)^2 ] / SIGMA(f)

STANDARD ERROR
SE = Population SD / SQRT(n)

CONFIDENCE INTERVAL (unknown SD)

CI = Mean +/- t * (SD / SQRT(n))

REGRESSION
Balance = -17732 + 367*Age + 1300*Education + 0.116*Wealth

RELATIVE FREQUENCY
Rel Freq = Class Count / Total Count

CLASS MIDPOINT
Midpoint = (Lower + Upper) / 2

Key Facts to Remember

Fact Value / Answer

Sum of deviations SIGMA(x-mu) Always = 0. Use this to verify your mean.

95% of data falls within Mean +/- 2 sigma (Two sigma rule)

Most common probabilistic method Simple Random Sampling

Subjective sampling method Judgment Sampling

Most used visualization software Tableau

z-test used when Population variance is KNOWN

t-test used when Population variance is UNKNOWN

 Type I Error Rejecting a TRUE null hypothesis (false alarm)

Type II Error Keeping a FALSE null hypothesis (missed detection)

Power of the test 1 - beta

Census vs Sample Census = ALL data. Sample = SUBSET of population.

Class width formula Lower bound of next class - lower bound of current class

Excel vertical bars Called Column charts (NOT bar charts)

Excel horizontal bars Called Bar charts

6-Step SD/Variance Quick Checklist

Find Mean (mu = SIGMA x / N)
1
Find deviations: (x - mu) for each value
2
Square each deviation: (x - mu)^2
3
Sum all squared deviations: SIGMA(x-mu)^2
4
Divide by N -> Variance (sigma^2)
5
Take SQRT -> Standard Deviation (sigma)
6

Exam Tip: Golden Rule: If SIGMA(deviations) is not 0, your mean is wrong — stop and recalculate before
continuing. All other steps depend on a correct mean.