Normal Distribution: Properties, Applications & Real-World Uses

TAKEAWAYS:

• The normal distribution is a symmetric, bell-shaped curve defined by its mean (μ) and standard deviation (σ), where 68% of data falls within ±1σ, 95% within ±2σ, and 99.7% within ±3σ.
• It is parametric (requires assumptions about μ and σ) and asymptotic (never touches the x-axis), making it ideal for modeling continuous data like heights, test scores, or financial returns.
• Applications include hypothesis testing (e.g., NTC’s call-center performance), quality control (e.g., Daraz’s product defect rates), and risk assessment (e.g., Nepal Bank’s loan defaults).
• The empirical rule (68-95-99.7) and z-scores ($z=\frac{X-\mu }{\sigma }$) are critical for converting raw data into probabilities and percentiles.
• Skewed data cannot be modeled by the normal distribution; use log transformations or non-parametric tests instead.
• Exam focus: Expect questions on calculating probabilities, finding percentiles, standardizing data, and interpreting real-world scenarios (e.g., "What % of Ncell customers spend > Rs. 5000/month?").

---

1. What Is the Normal Distribution?

The normal distribution is a continuous probability distribution where:

• Data clusters symmetrically around the mean (μ).
• The standard deviation (σ) determines spread.
• The curve is bell-shaped, asymptotic (never touches the x-axis), and unimodal.

Key Properties:

---

2. The Standard Normal Distribution (Z-Scores)

To compare values from different normal distributions, we standardize them using z-scores: $$z=\frac{X-\mu }{\sigma }$$

• Z-table: Converts z-scores to probabilities (e.g., $P(Z<1.2)=0.8849$).
• Symmetry: $P(Z>1.2)=1-P(Z<1.2)=0.1151$.

Worked Example 1: Ncell Customer Spending Ncell observes that monthly spending (X) follows $N(\mu =2000,\sigma =500)$. What % of customers spend:

1. Between Rs. 1500 and Rs. 2500?
2. More than Rs. 3000?

Solution:

1. Convert to z-scores: $$z_{1500}=\frac{1500-2000}{500}=-1,z_{2500}=\frac{2500-2000}{500}=1$$ From the z-table: $$P(-1<Z<1)=P(Z<1)-P(Z<-1)=0.8413-0.1587=0.6826(68.26\%)$$
2. For $X>3000$: $$z_{3000}=\frac{3000-2000}{500}=2$$ $$P(Z>2)=1-P(Z<2)=1-0.9772=0.0228(2.28\%)$$ Answer: 68.26% spend Rs. 1500–2500; 2.28% spend > Rs. 3000.

---

3. Applications of the Normal Distribution

A. Finance: Loan Defaults (Nepal Bank)

Nepal Bank uses the normal distribution to model loan repayment delays (X ~ $N(6,2)$ months, where μ = 6 months, σ = 2 months).

• Question: What % of loans are delayed > 10 months?
• Solution: $$z=\frac{10-6}{2}=2$$ $$P(X>10)=P(Z>2)=0.0228(2.28\%)$$ Interpretation: Nepal Bank expects ~2.3% of loans to be delayed >10 months, helping them set risk reserves.

B. Quality Control: Daraz Product Defects

Daraz tracks defective orders (X ~ $N(5,1.5)$ defects per 1000 items, μ = 5, σ = 1.5).

• Question: What % of batches have < 2 defects?
• Solution: $$z=\frac{2-5}{1.5}=-2$$ $$P(X<2)=P(Z<-2)=0.0228(2.28\%)$$ Action: Daraz aims to reduce defects to < 2% of batches via supplier audits.

C. Public Policy: NTC Call Center Wait Times

NTC models customer wait times (X ~ $N(120,30)$ seconds, μ = 120s, σ = 30s).

• Question: What % of calls wait > 2 minutes (120s)?
• Solution: $$z=\frac{120-120}{30}=0$$ $$P(X>120)=0.5(50\%)$$ Improvement: NTC targets reducing σ to 20s to cut long waits.

---

4. When Not to Use the Normal Distribution

The normal distribution fails for:

1. Skewed data (e.g., income distributions, stock returns).
   • Fix: Use log transformation or non-parametric tests.
2. Discrete data (e.g., Poisson-distributed events like accidents).
   • Fix: Use Poisson distribution.
3. Bimodal/multimodal data (e.g., heights of men and women combined).
   • Fix: Split into subgroups.

---

5. Key Formulas and Shortcuts

---

6. Common Mistakes to Avoid

1. Assuming all data is normal: Always check with a histogram or Q-Q plot.
2. Misapplying the empirical rule: Only works for perfectly normal data.
3. Ignoring units: Always label $\mu$ and $\sigma$ with units (e.g., "Rs.", "seconds").

---

7. Worked Example: NEPSE Stock Returns

NEPSE’s daily returns (X) follow $N(\mu =0.5\%,\sigma =1.2\%)$.

• Question: What % of days have returns between -1% and +2%?
• Solution:
  1. Convert to z-scores: $$z_{-1}=\frac{-1-0.5}{1.2}=-1.25,z_2=\frac{2-0.5}{1.2}=1.25$$
  2. Look up probabilities: $$P(Z<1.25)=0.8944,P(Z<-1.25)=0.1056$$
  3. Calculate: $$P(-1<X<2)=0.8944-0.1056=0.7888(78.88\%)$$ Interpretation: On ~79% of days, NEPSE returns fall in this range.

---

In the Real World

1. Khalti Digital Payments

   • Idea Used: Normal distribution models transaction amounts to detect fraud.
   • How: Khalti flags transactions > 3σ from the mean (e.g., if $\mu =5000$, $\sigma =1000$, a Rs. 10,000 transaction triggers review).
   • Visual:

2. Pathao Driver Earnings

   • Idea Used: Normal distribution predicts daily earnings to optimize driver incentives.
   • How: Pathao assumes earnings $X\sim N(12000,2000)$. Drivers earning < Rs. 8000 (below $\mu -2\sigma$) get bonuses.
   • Calculation: $$z=\frac{8000-12000}{2000}=-2\Rightarrow P(X<8000)=2.28\%$$ Action: Pathao targets < 2.3% of drivers for support.

3. NTC Internet Speed Tests

   • Idea Used: Normal distribution evaluates speed consistency.
   • How: NTC tests speeds $X\sim N(50,5)$ Mbps. Speeds < 40 Mbps (below $\mu -2\sigma$) trigger investigations.
   • Visual:

---

Exam Tip

1. Always check if data is normal before applying formulas. Use histograms or Q-Q plots if given raw data.
2. Memorize the empirical rule:
   • ±1σ: 68%
   • ±2σ: 95%
   • ±3σ: 99.7%
3. For probability questions:
   • Draw the normal curve and shade the area you’re solving for.
   • Use the z-table for exact values.
4. Real-world questions (e.g., "What % of Ncell users...") require:
   • Identifying $\mu$ and $\sigma$.
   • Converting to z-scores.
   • Interpreting the result in context (e.g., "2.28% of users spend > Rs. 3000").
5. Common exam traps:
   • Skewed data: If the distribution is not normal, do not use z-scores.
   • Discrete vs. continuous: The normal distribution is for continuous data (e.g., heights, weights). Use binomial/Poisson for counts (e.g., "number of defects").

---

Final Note: The normal distribution is the most powerful tool in Business Statistics for predicting probabilities, setting benchmarks, and making data-driven decisions. Master z-scores, the empirical rule, and real-world applications (like Ncell, Nepal Bank, and NTC examples above), and you’ll ace this unit!