MTH 216 — PROBABILITY & STATISTICS

EXAM-FOCUSED MASTER STUDY GUIDE

Pokhara University | BE Programme | 2014–2025 Analysis

★★★ HIGH-YIELD • PATTERN-BASED • EXAM-READY ★★★

HOW TO USE THIS GUIDE

★★★ = Appeared 5+ times | ★★ = Appeared 3-4 times | ★ = Appeared 1-2 times |

Focus on ★★★ topics first!

This guide covers ALL 8 units of MTH 216 in exam-priority order. Each section contains: Key Concepts
→ Formulas → Tricks → Mistakes → Solved Examples → Practice Questions.

★★★ MOST REPEATED QUESTIONS (prepare these FIRST):

1. Refrigerator Model A vs B — CV comparison (Unit I)
2. PMF table X: 2,3,4,5,6,7 — find K, E(X), V(X) (Unit III)
3. IQ before/after training paired t-test (Unit VII)
4. Age vs Blood Pressure correlation + regression (Unit VIII)
5. Bayes theorem — factory spring machines (Unit II)
6. Taxi drivers Poisson distribution (Unit III)
7. Normal distribution — 15%/40% first class/fail (Unit IV)

UNIT I: INTRODUCTION & DESCRIPTIVE STATISTICS

★★★ Topic 1: Coefficient of Variation (CV) — MOST REPEATED TOPIC

CV is the ratio of Standard Deviation to Mean, expressed as a percentage. It measures relative
variability.

CV = (σ / x̄) × 100%
Lower CV = More Uniform/Consistent | Higher CV = More Variable

SOLVED EXAMPLE — Model A vs B (Appears EVERY year!)

Given: Two models with frequency data over life groups. Steps:
1. Find mean (x̄) for each model: x̄ = Σ(f·x) / Σf
 2. Find variance: σ² = Σf(x−x̄)² / Σf → Find SD = √σ²
3. CV_A = (SD_A / x̄_A) × 100 → CV_B = (SD_B / x̄_B) × 100
4. Lower CV = Greater uniformity (better model)

💡 TIP: Mnemonic: 'LOWER CV = LESS Variation = MORE Uniform'. Always STATE

which model is more uniform and WHY.

⚠ COMMON MISTAKE: Always use class midpoints (not boundaries) for x in

frequency tables. E.g., for class 0-2: midpoint = 1.

Combined Mean and Variance Formula

Combined Mean = (n1·x̄1 + n2·x̄2) / (n1+n2)

Combined Variance σ² = [n1(σ1²+d1²) + n2(σ2²+d2²)] / (n1+n2)

where d1 = x̄1 − x̄combined

★★ Topic 2: Ogive (Cumulative Frequency Curve)

Steps to construct ogive:
5. Make cumulative frequency table (CF)
6. Plot Upper class boundary (x-axis) vs CF (y-axis)
7. Join points with smooth curve
8. To find value: draw horizontal line from CF-axis → hit curve → drop to x-axis

💡 TIP: For 'less than X marks': read CF directly. For 'between a and b': CF(b) − CF(a).
For 'more than x': Total − CF(x).

★ Topic 3: Stem-and-Leaf & Box Plot

Box Plot 5-number summary:
• Minimum, Q1 (lower quartile), Median (Q2), Q3 (upper quartile), Maximum
• IQR = Q3 − Q1
• Outlier if value < Q1 − 1.5×IQR or > Q3 + 1.5×IQR
• Skewed Right: Median closer to Q1 | Skewed Left: Median closer to Q3

Practice Questions — Unit I

Q1 [★★★]: Life of Model A/B refrigerators. Life groups 0-2,2-4,4-6,6-8,8-10,10-12 yrs; A:
5,16,13,7,5,4; B: 2,7,12,19,9,1. Find average life and CV. Which is more uniform?
Q2 [★★]: Firm A (n=500, mean=186, var=81) vs Firm B (n=600, mean=175, var=100). Find
combined average and variance.
 Q3 [★]: Construct ogive for: Marks 10-20,20-40,40-50,50-60,60-80,80-100; Students
50,100,150,90,60,50. Find: failed (below 40), top 25%.

UNIT II: PROBABILITY

★★★ Topic 1: Bayes' Theorem — FACTORY MACHINES (Appears 4+

times!)
Bayes' Theorem finds the reverse (posterior) probability:
P(Ai|B) = P(Ai) × P(B|Ai) / Σ P(Aj) × P(B|Aj)

STEP-BY-STEP METHOD — Spring Factory Problem

Given: Machine X (25% production, 5% defective), Y (35%, 4%), Z (40%, 2%)

Machine P(Machine) = Prior P(Defective|Machine) P(Machine) ×

P(Defective|Machine)
X 0.25 0.05 0.25 × 0.05 = 0.0125
Y 0.35 0.04 0.35 × 0.04 = 0.0140
Z 0.40 0.02 0.40 × 0.02 = 0.0080
TOTAL 1.00 — Σ = 0.0345

9. P(Defective) = Σ = 0.0345
10. P(Y|Defective) = 0.0140 / 0.0345 = 0.4058

💡 TIP: BAYES TABLE METHOD: Always make a 4-column table: Machine | Prior |
Likelihood | Joint. Sum the Joint column = Total probability. Divide each Joint by Total
= Posterior.

⚠ COMMON MISTAKE: Don't forget to STATE Bayes' theorem formula before applying
it. Examiners award marks for the formula statement.

★★★ Topic 2: Problem Given to Students A, B, C

Chances: P(A)=1/2, P(B)=3/4, P(C)=2/3 (or similar fractions). Since they work independently:
• P(all solve) = P(A)×P(B)×P(C)
• P(none solve) = P(A')×P(B')×P(C') where P(A') = 1 − P(A)
• P(problem solved) = 1 − P(none solve)
• P(exactly one) = P(A)·P(B')·P(C') + P(A')·P(B)·P(C') + P(A')·P(B')·P(C)
 💡 TIP: 'Problem can be solved' = at least one solves it = 1 − P(none). NEVER
calculate all cases individually; use complement!

★★ Topic 3: Bayes' Theorem — Students (Intelligence)

Classic: 75 students. Very intelligent=15, Medium=45, Below Average=15.
P(fail|very intelligent)=0.005, P(fail|medium)=0.05, P(fail|below avg)=0.15
Find P(below avg | passed) → Same Bayes table method, but event B = 'passed'.
💡 TIP: P(pass|category) = 1 − P(fail|category). Use Bayes table with 'pass' as the
event.

★ Topic 4: Mathematical Expectation

E(X) = Σ x·P(x) | E(aX+b) = a·E(X)+b | Var(aX+b) =
a²·Var(X)
Expectation of number of tosses until first head (Geometric):
E(X) = 1/p → For fair coin, p=1/2, so E(X) = 2

Practice Questions — Unit II

Q1 [★★★]: Factory — machines I,II,III produce 35%,25%,40%; defect rates 1%,2%,3%. Find
P(defective) and P(machine III | defective).
Q2 [★★★]: A,B,C solve prob with chances 1/2, 3/4, 2/3. Find P(all solve), P(none), P(at least
one), P(exactly one).
Q3 [★★]: 75 students — find P(below average | passed exam).
Q4 [★]: Coin tossed until head appears. Find expected number of tosses.

UNIT III: DISCRETE PROBABILITY DISTRIBUTIONS

★★★ Topic 1: PMF Table — Find K, E(X), V(X)

Given: X: 2,3,4,5,6,7 | P(X): 0.1, K, 0.2, 2K, 0.3, K

Step-by-Step Solution
11. Step 1 — Find K: Since ΣP(x) = 1: 0.1 + K + 0.2 + 2K + 0.3 + K = 1 → 0.6 + 4K = 1 → K =
0.1
12. Step 2 — Fill table: P(X): 0.1, 0.1, 0.2, 0.2, 0.3, 0.1
13. Step 3 — E(X) = Σ x·P(x) = 2(0.1)+3(0.1)+4(0.2)+5(0.2)+6(0.3)+7(0.1) = 4.8
14. Step 4 — E(X²) = Σ x²·P(x) = 4(0.1)+9(0.1)+16(0.2)+25(0.2)+36(0.3)+49(0.1) = 25.2
15. Step 5 — Var(X) = E(X²) − [E(X)]² = 25.2 − (4.8)² = 25.2 − 23.04 = 2.16
 E(aX+b) = a·E(X) + b | Var(aX+b) = a²·Var(X)

💡 TIP: Var(X) SHORTCUT: Always use Var(X) = E(X²) − [E(X)]². NEVER compute
Σ[x−μ]²·P(x) individually — it takes too long!

⚠ COMMON MISTAKE: Var(aX+b) = a²·Var(X) — the constant b disappears in variance!

Students often add b² by mistake.

★★★ Topic 2: Poisson Distribution — Taxi Drivers

Formula:
P(X=x) = e^(-λ) · λˣ / x! where λ = mean, e ≈ 2.71828

Given: λ = 3, n = 1000 drivers. e^(-3) = 0.0498

Requirement Formula Calculation Expected drivers
No accidents (x=0) e^(-3)·3⁰/0! 0.0498×1 = 0.0498 1000×0.0498 =
49.8 ≈ 50
Less than 2 (x=0,1) P(0)+P(1) 0.0498+0.1494 = 0.1992 1000×0.1992 =
199.2 ≈ 199
More than 3 1−P(0)−P(1)−P(2)−P(3) 1−0.0498−0.1494−0.224−0.224 1000×0.353 = 353
= 0.353

💡 TIP: POISSON KEY VALUES (memorize for λ=3): e^(-3)=0.0498, P(0)=0.0498,

P(1)=0.1494, P(2)=0.2240, P(3)=0.2240. 'More than 3' = 1 − sum of first 4.

⚠ COMMON MISTAKE: 'At least 3' means P(X≥3) = 1 − P(0) − P(1) − P(2). 'More than 3'
means P(X>3) = 1 − P(0) − P(1) − P(2) − P(3). These are DIFFERENT!

★★ Topic 3: Binomial Distribution

P(X=x) = C(n,x) · pˣ · (1-p)^(n-x) Mean=np Var=np(1-p)

Key identities:
• C(n,x) = n! / [x!(n-x)!]
• 'At least 1' = 1 − P(X=0) = 1 − (1−p)ⁿ
• 'At most 2' = P(0) + P(1) + P(2)

💡 TIP: When p is NOT given: 'succeeds twice as often as fails' means p=2/3, q=1/3.
'Boys and girls equal prob' means p=0.5.
Practice Questions — Unit III
Q1 [★★★]: X: −2,−1,0,1,2 | P(x): 0.2,0.1,0.3,0.3,0.1. Find E(X), E(2X−3), V(X), V(2X−3).
Q2 [★★★]: Poisson λ=3, n=1000 taxi drivers. Find expected with (i) >3 accidents (ii) <2
accidents (iii) no accidents.
Q3 [★★]: Experiment succeeds twice as often as fails. In 6 trials find P(at least 4
successes).
Q4 [★]: Binomial: n=5 trials, P(1 success)=0.4096, P(2 successes)=0.2048. Find p.

UNIT IV: CONTINUOUS PROBABILITY DISTRIBUTIONS

★★★ Topic 1: Normal Distribution — Find Mean & SD from Percentages

This is the most repeated type in Unit IV. The trick: convert % to z-score using the standard normal
table.

z = (X − μ) / σ → X = μ + z·σ

SOLVED EXAMPLE: 15% got first class (marks ≥ 60), 40% failed (marks < 40)
16. 15% got ≥ 60: P(X ≥ 60) = 0.15 → P(Z ≥ z1) = 0.15 → z1 = 1.04
17. So: 60 = μ + 1.04σ ... (equation 1)
18. 40% failed (< 40): P(X < 40) = 0.40 → P(Z < z2) = 0.40 → z2 = −0.25
19. So: 40 = μ − 0.25σ ... (equation 2)
20. Subtract eq2 from eq1: 20 = 1.29σ → σ = 15.50
21. From eq2: μ = 40 + 0.25(15.50) = 43.88

KEY Z-VALUES TO MEMORIZE (Standard Normal Table):

Area (one tail) Area (two tail = confidence) z-value

0.005 99% 2.576
0.025 95% 1.96
0.05 90% 1.645
0.10 80% 1.28
0.15 70% 1.04
0.20 60% 0.84
0.25 50% 0.675
0.30 40% 0.52
0.40 20% 0.25
 💡 TIP: PATTERN: Always set up TWO equations with μ and σ. The z-score for left tail
is NEGATIVE (use −z). Solve simultaneously.

⚠ COMMON MISTAKE: When problem says '25% scored MORE than 80' →
P(X>80)=0.25 → z is POSITIVE (+0.675). When '30% scored LESS than 45' →
P(X<45)=0.30 → z is NEGATIVE (−0.52).

★★ Topic 2: Exponential Distribution

f(x) = λe^(-λx) P(X ≤ x) = 1 − e^(-λx) Mean = 1/λ
Var = 1/λ²

EXAMPLE: Camera resets in < 20 days if λ = 1/50

22. P(X < 20) = 1 − e^(−(1/50)×20) = 1 − e^(-0.4)
23. e^(-0.4) ≈ 0.6703
24. P(X < 20) = 1 − 0.6703 = 0.3297

★★ Topic 3: Continuous PDF — Find k, E(X), Probabilities

For f(x) = kx² (0 < x < 1):
25. Find k: ∫₀¹ kx² dx = 1 → k[x³/3]₀¹ = 1 → k/3 = 1 → k = 3
26. E(X) = ∫₀¹ x·f(x)dx = ∫₀¹ 3x³dx = 3[x⁴/4]₀¹ = 3/4 = 0.75
27. P(0<x<0.7) = ∫₀^0.7 3x²dx = [x³]₀^0.7 = (0.7)³ = 0.343

💡 TIP: For ANY pdf: (1) Integrate and set = 1 to find k. (2) E(X) = ∫x·f(x)dx. (3)
P(a<X<b) = ∫ₐᵇ f(x)dx.

Practice Questions — Unit IV

Q1 [★★★]: In exam, 15% candidates got first class (≥60 marks), 40% failed (<40 marks).
Find mean and SD (normal distribution).
Q2 [★★★]: In exam, 25% scored >80, 30% scored <45. Find mean and SD.
Q3 [★★]: Normal dist, mean=73, SD=8. P(grade<91)=? Top 5% scored above what grade?
Q4 [★★]: f(x)=kx² for 0<x<1. Find k, E(X), P(0<x<0.7).
Q5 [★]: Camera, λ=1/50. P(reset in <20 days)=?

UNIT V: BIVARIATE RANDOM VARIABLES & JOINT

DISTRIBUTIONS
★★ Topic 1: Joint PMF — Marginal and Conditional Distributions
Given Joint PMF table with X and Y values:
Y=1 Y=2 Y=3 Marginal P(X)
X=2 0.10 0.20 0.10 0.40
X=3 0.15 0.30 0.15 0.60
Marginal P(Y) 0.25 0.50 0.25 1.00

• Marginal P(X=2) = sum of row X=2 = 0.10+0.20+0.10 = 0.40

• Marginal P(Y=1) = sum of col Y=1 = 0.10+0.15 = 0.25
• P(Y=y | X=3) = P(X=3,Y=y) / P(X=3) — divide each cell in row X=3 by 0.60
• X and Y are independent if P(X,Y) = P(X)·P(Y) for ALL combinations

★★ Topic 2: Joint PDF — Marginal, Conditional, Independence

For f(x,y) = 4xy, 0<x<1, 0<y<1:
28. Verify PDF: ∫₀¹∫₀¹ 4xy dxdy = 4·(1/2)·(1/2) = 1 ✓
29. Marginal f_X(x) = ∫₀¹ 4xy dy = 4x[y²/2]₀¹ = 2x
30. Marginal f_Y(y) = ∫₀¹ 4xy dx = 2y (by symmetry)
31. Check independence: f_X(x)·f_Y(y) = 2x·2y = 4xy = f(x,y) → X and Y ARE independent

💡 TIP: Independence shortcut: If f(x,y) can be written as g(x)·h(y), then X and Y are
independent. For f=4xy: 4xy = (2x)(2y) → independent!

⚠ COMMON MISTAKE: For f(x,y) = 2−x−y: cannot be factored → NOT independent.

Always verify by checking if f(x,y) = f_X(x)·f_Y(y).

Conditional PDF
f(x|y) = f(x,y) / f_Y(y) f(y|x) = f(x,y) / f_X(x)

Practice Questions — Unit V

Q1 [★★]: Joint PMF table X=(2,3), Y=(1,2,3): find marginals, P(Y|X=3), Var(X), Var(Y).
Q2 [★★]: f(x,y)=4xye^(-(x²+y²)), x>0,y>0. Test independence. Find conditional density
X|Y=y.
Q3 [★★]: f(x,y)=Axy, 0<x<1, 0<y<x. Find A, marginals, conditional f(y|x), check
independence.
Q4 [★]: f(x,y)=k(2x+y), 0≤x≤2, 0≤y≤3. Find marginals and check independence.

UNIT VI: SAMPLING DISTRIBUTION & ESTIMATION

★★★ Topic 1: Confidence Interval for Population Mean
CI = x̄ ± z_{α/2} · (σ/√n) [when σ known]
CI = x̄ ± t_{α/2, n-1} · (s/√n) [when σ unknown, use t-
distribution]

Confidence Level z-value

90% 1.645
95% 1.96
99% 2.576

SOLVED EXAMPLE — Light Bulb Factory (★★★ Repeated 3+ times!)

Given: σ=100 hrs, n=64, x̄=350 hrs. Find 95% CI.
32. SE = σ/√n = 100/√64 = 100/8 = 12.5
33. 95% CI: z = 1.96
34. Lower = 350 − 1.96×12.5 = 350 − 24.5 = 325.5
35. Upper = 350 + 1.96×12.5 = 350 + 24.5 = 374.5
36. Answer: 95% CI = (325.5, 374.5) hours

💡 TIP: SE (Standard Error) = σ/√n OR s/√n. ALWAYS calculate SE first, then multiply
by z or t. Width of CI = 2 × z × SE.

⚠ COMMON MISTAKE: Use z-table (1.96) when σ is KNOWN or n>30. Use t-table when
σ is UNKNOWN and n is small. Many students confuse these!

★★ Topic 2: Unbiased Estimator Proof

Steps for 'prove sample mean is unbiased estimator of population mean':
37. List all possible samples of size n from population (without replacement if stated)
38. Calculate mean of each sample
39. Find E(x̄) = Average of all sample means = ΣP(sample)·x̄(sample)
40. Show E(x̄) = μ (population mean) → sample mean is unbiased

💡 TIP: For population {4,2,0,6,8}: μ = (4+2+0+6+8)/5 = 4. For all C(5,2)=10 samples of

size 2, compute each mean. Average of those 10 means will equal 4.

★ Topic 3: Central Limit Theorem

CLT: As n→∞, sampling distribution of x̄ approaches Normal with μ_x̄ = μ and σ_x̄ =
σ/√n, REGARDLESS of population distribution.
Short Note: Properties of a Good Estimator (★★★ Short Note Topic!)
Mnemonic: 'SURE' — Sufficiency, Unbiasedness, Relative Efficiency, Efficiency
• Unbiasedness: E(estimator) = parameter
• Consistency: Estimate approaches true value as n→∞
• Efficiency: Minimum variance among all unbiased estimators
• Sufficiency: Uses all information in the sample

Practice Questions — Unit VI

Q1 [★★★]: σ=100, n=64, x̄=350. Find SE and 95% CI.
Q2 [★★]: Population {4,2,0,6,8}. Prove sample mean of size 2 is unbiased.
Q3 [★★]: Batch 2000, n=100, x̄=0.354, s=0.048. Find 95% CI.
Q4 [★]: n=300 students, mean=50, SD=10. Find 99% CI.

UNIT VII: HYPOTHESIS TESTING

★★★ Topic 1: Paired t-Test — IQ Training (Appears 6+ times — MOST

REPEATED!)
Given: IQ before and after training for 5 persons. Test if training is effective at 5% level.

Candidate Before (X₁) After (X₂) d = X₂−X₁ d²

1 110 120 10 100
2 120 118 −2 4
3 123 125 2 4
4 132 136 4 16
5 125 121 −4 16
Σ — — Σd=10 Σd²=140

41. d̄ = Σd/n = 10/5 = 2

42. Sd² = [Σd² − (Σd)²/n] / (n−1) = [140 − 100/5] / 4 = [140−20]/4 = 30
43. Sd = √30 = 5.477
44. t_calc = d̄ / (Sd/√n) = 2 / (5.477/√5) = 2 / 2.449 = 0.817
45. t_table at α=0.05, df=n−1=4 (two-tailed): t = 2.776
46. Since |t_calc| = 0.817 < t_table = 2.776: Fail to reject H₀
47. CONCLUSION: Training is NOT significantly effective at 5% level.

t = d̄ / (Sd/√n) where Sd = √[{Σd² − (Σd)²/n} / (n−1)]

 💡 TIP: PAIRED t-TEST STEPS: (1) Find d=after−before for each. (2) Find d̄ and Σd². (3)
Compute Sd. (4) Compute t_calc. (5) Compare with t_table (df=n−1). (6) State
conclusion.

⚠ COMMON MISTAKE: Use 'after−before' OR 'before−after' consistently. For training

problems, 'd=after−before'. If training is effective, we expect d̄ > 0. Hypotheses: H₀:
d̄=0 vs H₁: d̄≠0 (or >0 if one-tailed).

★★ Topic 2: Chi-Square Test of Independence

χ² = Σ (O−E)² / E where E = (Row Total × Column Total) /
Grand Total

Steps for Chi-Square Test

48. State H₀: Two variables are independent
49. Calculate expected frequencies: E = (RT × CT) / N
50. Compute χ² = Σ(O−E)²/E
51. df = (r−1)(c−1) for r rows and c columns
52. Compare χ²_calc with χ²_table at given α and df
53. If χ²_calc > χ²_table: Reject H₀ (variables are NOT independent)

💡 TIP: χ² critical values to remember: df=1 → 3.841 (5%), df=2 → 5.991 (5%), df=4 →
9.488 (5%). Always check if all E ≥ 5.

★★ Topic 3: Two-Sample z-Test for Means

z = (x̄₁ − x̄₂) / √(σ₁²/n₁ + σ₂²/n₂)
Example: Morning (n=200, x̄=72.4, s=14.8) vs Evening (n=100, x̄=73.9, s=17.9)
54. z = (72.4−73.9) / √(14.8²/200 + 17.9²/100) = −1.5/√(1.0952+3.2041) = −1.5/√4.299 = −0.724
55. z_table (5%, two-tail) = 1.96. |z_calc|=0.724 < 1.96 → No significant difference.

★ Topic 4: One-Sample t-Test (Small Sample)

t = (x̄ − μ₀) / (s/√n) df = n−1

Short Note: Type I and Type II Errors (★★★ Short Note Topic!)
Type I Error (α) Type II Error (β)
Reject H₀ when H₀ is TRUE Accept H₀ when H₀ is FALSE
False Positive False Negative
 = Level of Significance Power = 1−β
Controlled by researcher (usually 0.05) Related to sample size

💡 TIP: Memory: 'Type I = crying wolf (alarm when no danger). Type II = missing the
wolf (no alarm when danger exists).'

Practice Questions — Unit VII

Q1 [★★★]: IQ test before/after training (5 persons: 110,120,123,132,125 →
120,118,125,136,121). Is training effective at 5%?
Q2 [★★★]: Memory training: before (56,42,36,47,49,42,60,72,63), after
(147,125,118,128,145,140,155,160,149). Test at 5%.
Q3 [★★]: Hypertension vs smoking habits (contingency table). Chi-square test at α=0.05.
Q4 [★★]: Morning (n=200, x̄=72.4, s=14.8) vs Evening (n=100, x̄=73.9, s=17.9). Equal
means at 5%?
Q5 [★]: Factory fans: 2% defective in 3000 vs 3% defective in 2500. Is difference significant
at 5%?

UNIT VIII: CORRELATION & REGRESSION

★★★ Topic 1: Age vs Blood Pressure (10 Women) — MOST REPEATED!

Data: X (age): 56,42,36,47,49,42,60,72,63,55 | Y (BP): 147,125,118,128,145,140,155,160,149,150

Step-by-Step Solution
56. Calculate n=10, ΣX, ΣY, ΣXY, ΣX², ΣY²
57. For this data: ΣX=522, ΣY=1417, ΣXY=74,770, ΣX²=28,148, ΣY²=201,753
58. x̄=52.2, ȳ=141.7

Pearson's Correlation Coefficient

r = [ΣXY − n·x̄·ȳ] / √[(ΣX² − n·x̄²)(ΣY² − n·ȳ²)]

Or equivalently:
r = [nΣXY − ΣX·ΣY] / √[(nΣX² − (ΣX)²)(nΣY² − (ΣY)²)]

Regression Line Y on X
ŷ = a + bx where b = r·(Sy/Sx) OR b = [ΣXY − n·x̄·ȳ] /
[ΣX² − n·x̄²]
 a = ȳ − b·x̄

Regression Line X on Y
x̂ = a' + b'y where b' = r·(Sx/Sy) OR b' = [ΣXY −
n·x̄·ȳ] / [ΣY² − n·ȳ²]

★★ Topic 2: Coefficient of Determination and Standard Error

R² = r² (proportion of variation in Y explained by X)
Se = √[ΣY² − a·ΣY − b·ΣXY] / (n−2)

💡 TIP: QUICK CHECK: r should be between −1 and +1. If |r| > 0.8 = strong, 0.5-0.8 =
moderate, < 0.5 = weak. For Age-BP data, r ≈ 0.9 (strong positive).

⚠ COMMON MISTAKE: Regression of Y on X and X on Y are DIFFERENT lines. Y on X

is used to predict Y from X. X on Y is used to predict X from Y. Don't mix them up!

Properties of Correlation Coefficient (Short Note ★★)

• r is dimensionless (no units)
• −1 ≤ r ≤ +1
• r = +1: perfect positive linear relationship
• r = −1: perfect negative linear relationship
• r = 0: no linear relationship (may have non-linear)
• r is symmetric: r(X,Y) = r(Y,X)
• r is not affected by change of origin and scale

Practice Questions — Unit VIII

Q1 [★★★]: Age X and BP Y of 10 women (data above). Find r and regression equation Y on
X.
Q2 [★★]: Forestry data: n=6, ΣX=21, ΣY=26, ΣX²=91, ΣY²=142.52, ΣXY=113.8. Find
regression, r, R², Se.
Q3 [★★]: Marks in Maths (85,60,73,40,82,90) and Physics (92,72,75,65,50,80). Regress
Physics on Maths. Predict Physics for Maths=52.
Q4 [★]: Write properties of correlation and regression coefficients.

SHORT NOTES MASTER REFERENCE (Q7 — Any Two)

★★★ Most Likely Short Note Topics for Exam

GUARANTEED SHORT NOTES — Must prepare all of these:

1. Properties / Criteria of Good Estimator (★★★ Appeared 10+ times!)

Mnemonic: UCES — Unbiased, Consistent, Efficient, Sufficient
• Unbiasedness: E(θ̂) = θ (estimator hits the target on average)
• Consistency: As n→∞, θ̂→θ (more data = closer to truth)
• Efficiency: Minimum variance among unbiased estimators (most precise)
• Sufficiency: Uses ALL information available in sample

2. Type I and Type II Errors in Hypothesis Testing (★★★)

Error Type Description
Type I Error (α) Reject H₀ when H₀ is TRUE (False alarm).
Controlled by significance level.
Type II Error (β) Accept H₀ when H₀ is FALSE (Miss). Power =
1−β.
Relationship Reducing α increases β and vice versa for fixed n.
Solution Increase sample size to reduce both errors
simultaneously.

3. Probability and Non-Probability Sampling (★★★)

Probability Sampling Non-Probability Sampling
Random selection Non-random selection
Each element has known probability Probability unknown
Simple Random, Stratified, Cluster, Systematic Convenience, Judgment, Quota, Snowball
Unbiased; representative May be biased; faster/cheaper

4. Binomial Distribution (★★)

• Conditions (FINES): Fixed n trials, Independent trials, N constant (n fixed), Either success or
failure, Same probability p each trial
• PMF: P(X=x) = C(n,x)·pˣ·qⁿ⁻ˣ where q=1−p
• Mean = np, Variance = npq, SD = √(npq)
• As n→∞ and p→0 with np=λ constant: Binomial → Poisson

5. Normal Distribution and its Characteristics (★★)

• Bell-shaped, symmetric about mean μ
• Mean = Median = Mode = μ
 • Total area under curve = 1
• μ ± σ covers 68.27% data
• μ ± 2σ covers 95.45% data
• μ ± 3σ covers 99.73% data
• Points of inflection at μ ± σ
• Skewness = 0, Kurtosis = 3

6. Coefficient of Determination R² (★★)

• R² = r² = (explained variation) / (total variation)
• R² ranges from 0 to 1 (or 0% to 100%)
• R²=0.81 means 81% of variation in Y is explained by X
• Remaining (1−R²) is unexplained variation

7. Central Limit Theorem (★★)

'As sample size n increases (n>30), the sampling distribution of the sample mean approaches a normal
distribution with mean μ and standard deviation σ/√n, regardless of the shape of the population
distribution.'
• Importance: Allows use of normal distribution for inference even for non-normal populations
• z = (x̄ − μ) / (σ/√n) is approximately standard normal for large n

8. Confidence Interval / Interval Estimation (★★)

• Point estimate: Single value (e.g., x̄ = 50)
• Interval estimate: Range with confidence level (e.g., 45 < μ < 55 at 95%)
• Width decreases with: larger n, smaller σ, lower confidence level
• Interpretation: '95% CI means if we repeated sampling 100 times, 95 CIs would contain true μ'

9. Cluster vs Stratified Sampling (★)

Cluster Sampling Stratified Sampling
Population divided into clusters Population divided into strata (groups)
Select ENTIRE clusters randomly Select SAMPLES from each stratum
Clusters = heterogeneous inside Strata = homogeneous inside
Used when groups are geographically spread Used when population has distinct sub-groups
Less expensive, less precise More expensive, more precise

QUICK REVISION: MEGA FORMULA SHEET

All Key Formulas — One-Page Reference

Descriptive Statistics
Mean: x̄ = Σfx/Σf SD: σ = √[Σf(x−x̄)²/Σf] CV =
(σ/x̄)×100%

Probability
P(A∪B) = P(A)+P(B)−P(A∩B) P(A∩B) = P(A)·P(B|A)
Bayes: P(Ai|B) = P(Ai)·P(B|Ai) / Σ P(Aj)·P(B|Aj)

Expectation and Variance

E(X)=Σx·P(x) Var(X)=E(X²)−[E(X)]² E(aX+b)=aE(X)+b
Var(aX+b)=a²Var(X)

Discrete Distributions
Binomial: P(x)=C(n,x)·pˣ·qⁿ⁻ˣ Mean=np Var=npq
Poisson: P(x)=e^(-λ)·λˣ/x! Mean=λ Var=λ

Continuous Distributions
Normal: z=(X−μ)/σ Exponential: P(X≤x)=1−e^(-λx) Mean=1/λ

Confidence Intervals
σ known: CI = x̄ ± z·(σ/√n) σ unknown: CI = x̄ ± t_{n-
1}·(s/√n)

Hypothesis Testing
Paired t: t = d̄/(Sd/√n) Sd=√[{Σd²−(Σd)²/n}/(n−1)]
df=n−1

Two-sample z: z=(x̄₁−x̄₂)/√(σ₁²/n₁+σ₂²/n₂)
Chi-square: χ²=Σ(O−E)²/E E=(RT×CT)/N df=(r−1)(c−1)

Correlation and Regression

 r = [nΣXY−ΣX·ΣY] / √[(nΣX²−(ΣX)²)(nΣY²−(ΣY)²)]
Regression Y on X: ŷ=a+bx b=[nΣXY−ΣX·ΣY]/[nΣX²−(ΣX)²]
a=ȳ−b·x̄
R² = r² Se = √[ΣY²−a·ΣY−b·ΣXY]/(n−2)

EXAM DAY REMINDER: Always write the formula first, then substitute values, then
calculate. Show all steps clearly — partial marks are given!

For ANY question, start with: State the formula/theorem → Identify given values →
Apply step-by-step → State conclusion clearly.