AS & A Level Mathematics · Probability & Statistics 1 · Chapter 8
The normal distribution
- The Normal distribution: the most vital continuous probability model in statistics, governing natural variation, measurement errors, and sampling theory.
- Key properties: symmetrical bell-shaped curve, mean = median = mode = μ, total area = 1, and inflection points at μ ± σ.
- Core A-Level techniques: standardisation Z = (X − μ)/σ, standard normal table navigation, setting up simultaneous equations to solve for unknown μ and σ, and approximating binomial models using continuity corrections.
🎯 Learning objectives & Book exercises
- Understand the nature of continuous random variables and probability density functions (PDFs).
- Recognise that probability corresponds to area under the PDF curve: total area = 1, and P(X = c) = 0 for any single point c.
- Recognise the normal distribution X ~ N(μ, σ²), its bell shape, line of symmetry at x = μ, and points of inflection at x = μ ± σ.
- Understand the effect of altering the mean μ (horizontal translation) and standard deviation σ (spread and height scaling).
- Book exercises covered: Exercise 8A and matching review questions.
8.1 Continuous random variables & the normal curve
- Continuous random variable: a variable that can take any real value, typically over an interval or continuum (such as time, mass, distance, voltage, or temperature).
- Probability density function (PDF): a non-negative function f(x) representing the relative likelihood of outcomes:
- f(x) ≥ 0 for all real values of x.
- The total area enclosed between y = f(x) and the x-axis is always exactly 1.
- Single point probability is zero: for any continuous random variable, P(X = c) = 0 because the area of a vertical line with zero width is 0.
- Inequality signs are interchangeable: because single points carry zero probability, strict and non-strict inequalities have identical values:
P(a < X < b) = P(a ≤ X < b) = P(a < X ≤ b) = P(a ≤ X ≤ b)
⭐ Key Point 8.1: The Normal Probability Density Function
- A random variable X is normally distributed with mean μ and variance σ²: written X ~ N(μ, σ²).
- Its mathematical equation is:
f(x) = exp[ − ], for all x ∈ ℝ
- Symmetry: the curve is perfectly symmetrical about the vertical axis x = μ. Hence, Mean = Median = Mode = μ.
- Inflection points: the curve changes curvature (from concave downwards to concave upwards) at exactly x = μ − σ and x = μ + σ.
- Asymptotes: the curve extends indefinitely in both directions, approaching the horizontal axis asymptotically without ever touching it.
Altering parameters: Location μ versus Dispersion σ
- Changing mean μ (with constant σ):
- The entire curve shifts horizontally without changing width or height.
- If μ₂ > μ₁, curve 2 is obtained by translating curve 1 to the right by (μ₂ − μ₁) units.
- Changing standard deviation σ (with constant μ):
- Standard deviation controls both the spread and the maximum height of the peak.
- Because the total area must remain 1: maximum height = .
- Smaller σ yields a narrower, taller peak (data clustered closely around μ).
- Larger σ yields a flatter, wider bell curve (data dispersed further from μ).
💡 Worked Example 8.A: Comparing Normal Distributions
The quantities of apple juice in 500 tins follow a normal distribution with mean 340 ml and variance 4 ml². The quantities of peach juice in 500 tins follow a normal distribution with mean 340 ml and standard deviation 4 ml. Describe the similarities and differences between the two normal curves.
- Step 1: Extract parameters for each distribution
For apple juice: mean μ_A = 340 ml, variance σ_A² = 4 ml² ⇒ standard deviation σ_A = √4 = 2 ml.
For peach juice: mean μ_P = 340 ml, standard deviation σ_P = 4 ml (so variance σ_P² = 16 ml²). - Step 2: Compare location (centres)
Both distributions have identical means: μ_A = μ_P = 340 ml. Both curves share the exact same line of symmetry at x = 340 ml and have equal medians and modes of 340 ml. - Step 3: Compare dispersion (spread and height)
Peach juice has twice the standard deviation of apple juice (σ_P = 2σ_A). Therefore, the peach juice curve is twice as spread out horizontally.
Because total area under both curves is 1, the peak height of the peach juice curve is exactly half the peak height of the apple juice curve ( vs ). - Step 4: State clear comparative conclusion
Answer: Both curves are symmetrical bells centered at 340 ml; the apple juice curve has a higher peak and narrower spread (inflection points at 338 and 342 ml), whereas the peach juice curve is flatter and wider (inflection points at 336 and 344 ml).
✏️ Try it: Continuous variables & normal curve
- A continuous random variable X has normal distribution X ~ N(50, 64). State the mean, the standard deviation, and the coordinates of the two points of inflection on the curve.
- Dataset A has totals Σx = 12000, Σx² = 35000, n = 5000. Dataset B has totals Σy = 26000, Σy² = 72000, n = 10000. Show that the centre of curve B is located to the right of curve A.
- For a normal variable X with mean 18 and standard deviation 3, write down the approximate percentage of observations falling in the interval [12, 24].
Answers
- 1. Step 1: μ = 50, variance = 64 ⇒ σ = √64 = 8. Step 2: Points of inflection occur at x = μ ± σ ⇒ x = 50 − 8 = 42 and x = 50 + 8 = 58. Step 3: Maximum peak is at (50, ). Answer: Mean = 50, σ = 8, inflection points at x = 42 and x = 58.
- 2. Step 1: Mean of A: x̄ = Σx / n = 12000 / 5000 = 2.4. Step 2: Mean of B: ȳ = Σy / n = 26000 / 10000 = 2.6. Step 3: Since ȳ = 2.6 > x̄ = 2.4, the centre of curve B is situated to the right of curve A.
- 3. Step 1: Note that 18 − 2(3) = 12 and 18 + 2(3) = 24. Step 2: The interval [12, 24] represents [μ − 2σ, μ + 2σ]. Step 3: By the empirical rule, approximately 95.45% (or ~95%) of observations fall within 2 standard deviations of the mean.
🎯 Learning objectives & Book exercises
- Understand the standard normal variable Z ~ N(0, 1) with mean 0 and variance 1.
- Master the use of Cambridge standard normal cumulative distribution tables for Φ(z) = P(Z < z).
- Apply reflection symmetry Φ(−z) = 1 − Φ(z) to evaluate left tails and negative z-scores.
- Calculate interval probabilities P(a < Z < b) = Φ(b) − Φ(a) and central intervals P(−a < Z < a) = 2Φ(a) − 1.
- Perform inverse lookups to determine critical values: finding z given P(Z < z) = p or P(Z > z) = p.
- Book exercises covered: Exercise 8B and matching review questions.
8.2 The standard normal variable Z ~ N(0, 1)
- Definition: the standard normal distribution is the special normal curve with mean μ = 0 and standard deviation σ = 1 (variance σ² = 1).
- Cumulative function Φ(z): the Greek capital letter Phi, Φ(z), denotes the probability that Z is less than z:
Φ(z) = P(Z < z) = P(Z ≤ z)
- The Cambridge normal table:
- Tables give values of Φ(z) for positive z from 0.000 to 2.999.
- The first two decimal digits are read from the left column and top row; the third decimal digit is added using the "ADD" differences column on the right.
- At z = 0: Φ(0) = 0.5000 (exactly half the area lies below the mean).
⭐ Key Point 8.2: Symmetry & The Four Standard Tail Rules
- 1. Left tail for positive z:
P(Z < z) = Φ(z), for z ≥ 0
- 2. Right tail for positive z (Complement):
P(Z > z) = 1 − P(Z ≤ z) = 1 − Φ(z)
- 3. Left tail for negative z (Reflection Symmetry):
By symmetry, the area in the left tail below −z equals the area in the right tail above +z:
Φ(−z) = P(Z < −z) = P(Z > z) = 1 − Φ(z)
- 4. Right tail for negative z:
P(Z > −z) = 1 − P(Z < −z) = 1 − (1 − Φ(z)) = Φ(z)
- 5. Interval between two values:
P(a < Z < b) = Φ(b) − Φ(a)
- 6. Symmetric central interval:
P(−a < Z < a) = Φ(a) − Φ(−a) = Φ(a) − (1 − Φ(a)) = 2Φ(a) − 1
💡 Worked Example 8.1: Finding Standard Normal Probabilities
Given that Z ~ N(0, 1), calculate each of the following probabilities correct to 3 significant figures:
(a) P(Z < 1.43)
(b) P(Z > 2.08)
(c) P(Z < −1.752)
- Step 1: Calculate (a) P(Z < 1.43) via direct table lookup
Because 1.43 is positive and we seek the area to the left, this is directly given by Φ(1.43).
In the tables, find row 1.4 and column 0.03 ⇒ Φ(1.43) = 0.9236.
Answer (a): 0.924 (to 3 s.f.). - Step 2: Calculate (b) P(Z > 2.08) using complement rule
P(Z > 2.08) = 1 − P(Z ≤ 2.08) = 1 − Φ(2.08).
From the tables: row 2.0, column 0.08 gives Φ(2.08) = 0.9812.
Compute: 1 − 0.9812 = 0.0188.
Answer (b): 0.0188 (to 3 s.f.). - Step 3: Calculate (c) P(Z < −1.752) using reflection symmetry
By symmetry: Φ(−1.752) = 1 − Φ(1.752).
In tables: row 1.7, column 0.05 gives 0.9599. The 3rd decimal is 2; in the 'ADD 2' column, the value is 2 (i.e. +0.0002).
So Φ(1.752) = 0.9599 + 0.0002 = 0.9601.
Compute: 1 − 0.9601 = 0.0399.
Answer (c): 0.0399 (to 3 s.f.).
💡 Worked Example 8.2: Interval Probabilities on Z
Find P(−1.00 < Z < 2.115) where Z ~ N(0, 1).
- Step 1: Express interval as the difference of two cumulative probabilities
P(−1.00 < Z < 2.115) = Φ(2.115) − Φ(−1.00). - Step 2: Look up Φ(2.115) using ADD column
Base value at z = 2.11 is 0.9826. In column 'ADD 5', the number is 2 (+0.0002).
So Φ(2.115) = 0.9826 + 0.0002 = 0.9828. - Step 3: Express negative z value using symmetry
Φ(−1.00) = 1 − Φ(1.00) = 1 − 0.8413 = 0.1587. - Step 4: Subtract and state final probability
P = 0.9828 − 0.1587 = 0.8241.
Answer: 0.824 (or 0.8241).
💡 Worked Example 8.3: Inverse Normal Lookup (Finding Critical Values)
Given that Z ~ N(0, 1), find the value of b such that P(Z ≥ b) = 0.7713.
- Step 1: Determine the sign of b from the probability
Since P(Z ≥ b) = 0.7713 > 0.5, more than half of the distribution lies to the right of b.
Therefore, b must lie to the left of the mean 0: b is negative. Let b = −a, where a > 0. - Step 2: Rephrase using reflection symmetry
P(Z ≥ −a) = P(Z ≤ a) = Φ(a) = 0.7713. - Step 3: Look up Φ⁻¹(0.7713) in the standard normal tables
Look for probability 0.7713 inside the body of the table.
The closest tabulated value below 0.7713 is 0.7704 at row 0.7, column 0.04 (z = 0.74).
Difference required: 0.7713 − 0.7704 = 0.0009.
In the 'ADD' section for row 0.7, difference 9 corresponds to column 3.
Therefore: a = 0.740 + 0.003 = 0.743. - Step 4: Restore negative sign
Since b = −a, we have b = −0.743.
Answer: b = −0.743. Check: P(Z ≥ −0.743) = Φ(0.743) = 0.7713.
✏️ Try it: Standard normal Z & tables
- Given Z ~ N(0, 1), find P(Z > 0.817) and P(Z < −1.75).
- Find P(−1.20 < Z < 1.20).
- Given that P(Z < k) = 0.3552, find the value of k.
Answers
- 1. Step 1: For P(Z > 0.817) = 1 − Φ(0.817). From tables: Φ(0.81) = 0.7910; ADD 7 gives +0.0020 ⇒ Φ(0.817) = 0.7930. 1 − 0.7930 = 0.207.
Step 2: For P(Z < −1.75) = 1 − Φ(1.75) = 1 − 0.9599 = 0.0401. - 2. Step 1: Central interval formula: P(−1.20 < Z < 1.20) = 2Φ(1.20) − 1. Step 2: From tables, Φ(1.20) = 0.8849. Step 3: Compute 2(0.8849) − 1 = 1.7698 − 1 = 0.770 (or 0.7698).
- 3. Step 1: Since P(Z < k) = 0.3552 < 0.5, k is negative: let k = −c. Step 2: P(Z < −c) = 1 − Φ(c) = 0.3552 ⇒ Φ(c) = 1 − 0.3552 = 0.6448. Step 3: From tables, 0.6443 at z = 0.37; difference 5 is in ADD 1. So c = 0.371. Step 4: k = −0.371.
🎯 Learning objectives & Book exercises
- Understand the standardisation principle: transforming any normal variable X ~ N(μ, σ²) into standard normal Z ~ N(0, 1).
- Use the standardisation formula Z = to compute z-scores.
- Calculate tail and interval probabilities for general normal distributions.
- Apply reverse standardisation: finding values of x corresponding to given probabilities using x = μ + zσ.
- Book exercises covered: Exercise 8C and matching review questions.
8.2 Standardising any normal distribution X ~ N(μ, σ²)
- The need for standardisation: an infinite number of normal curves exist, each with its own mean μ and variance σ². Standardisation maps every normal curve onto the single reference curve Z ~ N(0, 1).
- Coding mechanism:
- Subtracting μ shifts the entire distribution horizontally so that the new mean is 0: E(X − μ) = 0.
- Dividing by standard deviation σ rescales the spread so that the new variance is 1: Var((X − μ)/σ) = Var(X) = = 1.
- The z-score: a standardized score z represents the number of standard deviations that a value x lies above or below the mean μ.
⭐ Key Point 8.3: Standardisation Formulas
- Forward standardisation (finding probabilities):
z = ⇒ P(X < x) = P( Z < ) = Φ(z)
- Reverse standardisation (finding cutoff values):
Given probability P(X < x) = p: find z = Φ⁻¹(p) from tables, then rearrange:
x = μ + zσ
Note: If the percentile is below 50%, z is negative, so x lies below μ.
🕹️ Interactive Simulation: Normal Bell Curve & Probability Explorer
Adjust the mean μ and standard deviation σ sliders below, choose an interval type, and observe the live Gaussian bell curve, shaded probability region, and step-by-step z-score derivation.
💡 Worked Example 8.6: Standardising a Normal Distribution
The continuous random variable X follows the normal distribution X ~ N(6.2, 6.25).
Calculate (a) P(X ≤ 9.7) and (b) P(X > 9.7), correct to 3 significant figures.
- Step 1: Identify distribution parameters
Mean μ = 6.2, variance σ² = 6.25 ⇒ standard deviation σ = √6.25 = 2.5. - Step 2: Standardise x = 9.7 to obtain z-score
z = = = = 1.40. - Step 3: Evaluate P(X ≤ 9.7) using standard normal tables
P(X ≤ 9.7) = P(Z ≤ 1.40) = Φ(1.40).
From the normal distribution table: Φ(1.40) = 0.9192.
Answer (a): 0.919 (or 0.9192). - Step 4: Evaluate P(X > 9.7) using the complement rule
P(X > 9.7) = 1 − P(X ≤ 9.7) = 1 − 0.9192 = 0.0808.
Answer (b): 0.0808 (to 3 s.f.). Check: 0.9192 + 0.0808 = 1.0000.
💡 Worked Example 8.7: Reverse Standardisation (Finding Cutoff Value)
Given that X ~ N(30, 16) and that P(X ≤ a) = 0.8944, find the value of a.
- Step 1: Extract parameters
Mean μ = 30, variance σ² = 16 ⇒ standard deviation σ = √16 = 4. - Step 2: Standardise the probability statement
P( Z ≤ ) = 0.8944 ⇒ Φ( ) = 0.8944. - Step 3: Look up inverse normal Φ⁻¹(0.8944)
In the tables, probability 0.8944 appears exactly at z = 1.25.
So = 1.25. - Step 4: Solve for a and check
a − 30 = 4 × 1.25 = 5.0 ⇒ a = 30 + 5.0 = 35.
Answer: a = 35. Check: (35 − 30)/4 = 1.25, and Φ(1.25) = 0.8944.
✏️ Try it: Standardising X ~ N(μ, σ²)
- Given X ~ N(20, 15), calculate P(X < 13.5) correct to 3 significant figures.
- Given X ~ N(11, 25), find P(1 ≤ X < 21).
- Given X ~ N(100, 64) and P(X ≥ e) = 0.95, find the value of e.
Answers
- 1. Step 1: μ = 20, σ = √15 ≈ 3.873. Step 2: z = (13.5 − 20)/√15 = −6.5 / 3.873 = −1.678. Step 3: P(Z < −1.678) = 1 − Φ(1.678). Tables: Φ(1.67) = 0.9525, ADD 8 gives +0.0008 ⇒ Φ(1.678) = 0.9533. 1 − 0.9533 = 0.0467.
- 2. Step 1: μ = 11, σ = √25 = 5. Step 2: Standardise bounds: z₁ = (1 − 11)/5 = −10/5 = −2.00; z₂ = (21 − 11)/5 = 10/5 = 2.00. Step 3: Central interval: P(−2 ≤ Z < 2) = 2Φ(2.00) − 1 = 2(0.9772) − 1 = 1.9544 − 1 = 0.954 (or 0.9544).
- 3. Step 1: μ = 100, σ = 8. P(X ≥ e) = 0.95 ⇒ P(X < e) = 0.05. Step 2: e lies in the lower 5%, so z is negative. From tables: Φ(1.645) = 0.95 ⇒ z = −1.645. Step 3: (e − 100)/8 = −1.645 ⇒ e = 100 − 8(1.645) = 100 − 13.16 = 86.84 (or 86.8).
🎯 Learning objectives & Book exercises
- Apply the normal distribution to real-world continuous phenomena (lengths, masses, durations, tolerances).
- Calculate one unknown parameter (mean μ or standard deviation σ) given a single percentile condition.
- Set up and solve pairs of simultaneous linear equations to find both unknown μ and σ from two probability conditions.
- Solve two-stage problems combining normal distribution probabilities with binomial distributions.
- Book exercises covered: Exercise 8D and matching review questions.
8.3 Modelling with the normal distribution
- Physical rationale: Carl Friedrich Gauss demonstrated that random measurement errors accumulate additively, creating the characteristic bell-shaped curve.
- Single unknown parameter:
- If σ is known and μ is unknown: standardise the given condition P(X < x) = p to find z = Φ⁻¹(p), then solve μ = x − zσ.
- If μ is known and σ is unknown: standardise to find z, then solve σ = (x − μ) / z.
- Two unknown parameters (μ and σ both unknown):
- Two distinct probability conditions are required: P(X < x₁) = p₁ and P(X < x₂) = p₂ (with x₁ < x₂).
- Each condition translates to a standard z-score: z₁ = Φ⁻¹(p₁) and z₂ = Φ⁻¹(p₂).
- This yields two simultaneous linear equations:
x₁ − μ = z₁σ [Equation 1]
x₂ − μ = z₂σ [Equation 2]
- Subtracting Eq 1 from Eq 2 eliminates μ:
(x₂ − x₁) = (z₂ − z₁)σ ⇒ σ =
- Substitute σ back into Equation 1 to find: μ = x₁ − z₁σ.
🧮 Interactive Tool: Normal Parameters & Simultaneous Equations Solver
Enter two data points and their corresponding cumulative or tail probabilities to calculate the unknown mean μ and standard deviation σ with full algebraic working.
Condition 1
Condition 2
💡 Worked Example 8.9: Solving for Unknown μ and σ via Simultaneous Equations
The continuous random variable X is normally distributed with mean μ and variance σ². Given that P(X < 10) = 0.25 and P(X < 12) = 0.90, calculate the value of μ and of σ.
- Step 1: Standardise condition 1 and find z₁
P(X < 10) = 0.25 < 0.5. By symmetry, the 25th percentile lies below the mean.
From tables: Φ(0.674) = 0.7500 ⇒ z₁ = −0.674.
Standardisation equation: = −0.674 ⇒ 10 − μ = −0.674σ [Eq 1]. - Step 2: Standardise condition 2 and find z₂
P(X < 12) = 0.90 > 0.5. The 90th percentile lies above the mean.
From tables: Φ(1.282) = 0.9000 ⇒ z₂ = +1.282.
Standardisation equation: = +1.282 ⇒ 12 − μ = 1.282σ [Eq 2]. - Step 3: Subtract Eq 1 from Eq 2 to eliminate μ and solve for σ
(12 − μ) − (10 − μ) = 1.282σ − (−0.674σ)
2 = 1.956σ ⇒ σ = ≈ 1.0225 ⇒ 1.02 (to 3 s.f.). - Step 4: Substitute σ into Eq 2 to find μ
μ = 12 − 1.282(1.0225) = 12 − 1.3108 = 10.69 (or 10.7 to 3 s.f.). - Step 5: Verify with Eq 1
Check: 10 − (−0.674)(1.0225) = 10 + 0.689 = 10.69. Verified!
Answer: μ = 10.7, σ = 1.02 (or μ = 10.69, σ = 1.022).
💡 Worked Example 8.10: Two-Stage Problem (Normal into Binomial)
The volumes of oil in tins are normally distributed with mean 508 ml and standard deviation 15 ml.
(a) Find the probability that a randomly chosen tin contains less than 500 ml.
(b) A random sample of 3 tins is taken. Find the probability that exactly 1 tin contains less than 500 ml.
- Step 1: Normal stage — standardise x = 500
X ~ N(508, 15²).
z = = ≈ −0.533. - Step 2: Calculate tail probability using symmetry
P(X < 500) = P(Z < −0.533) = 1 − Φ(0.533).
From tables: Φ(0.53) = 0.7019; ADD 3 gives +0.0011 ⇒ Φ(0.533) = 0.7030.
P = 1 − 0.7030 = 0.2970.
Answer (a): 0.297 (or 29.7%). - Step 3: Binomial stage — define discrete variable Y
Let Y be the number of tins containing < 500 ml in a sample of n = 3.
Each tin is independent with p = 0.297. Hence, Y ~ B(3, 0.297). - Step 4: Calculate P(Y = 1) using binomial formula
P(Y = 1) = (0.297)¹ (1 − 0.297)² = 3 × 0.297 × (0.703)².
Compute: 3 × 0.297 × 0.494209 = 0.4403.
Answer (b): 0.440 (to 3 s.f.).
✏️ Try it: Modelling & finding unknown μ and σ
- The heights of trees in a forest are normally distributed with mean μ and standard deviation 3.6 m. Given that 75% of the trees are less than 10 m tall, find the value of μ.
- The speeds of cars are normally distributed with mean μ and standard deviation 20 km/h. Given that 33% of cars travel over 100 km/h, find μ and calculate the percentage traveling under 80 km/h.
- For a normal variable W ~ N(μ, σ²), P(W < 8.4) = 0.7509 and P(W > 9.2) = 0.1385. Find the value of μ and σ.
Answers
- 1. Step 1: P(X < 10) = 0.75. From tables: Φ(0.674) = 0.75 ⇒ z = 0.674. Step 2: (10 − μ)/3.6 = 0.674 ⇒ 10 − μ = 2.426. Step 3: μ = 10 − 2.426 = 7.57 m (or 7.574).
- 2. Step 1: P(S > 100) = 0.33 ⇒ P(S < 100) = 0.67. Tables: Φ(0.44) = 0.6700 ⇒ z = 0.440. Step 2: (100 − μ)/20 = 0.440 ⇒ 100 − μ = 8.80 ⇒ μ = 91.2 km/h. Step 3: P(S < 80): z = (80 − 91.2)/20 = −11.2/20 = −0.56. P(Z < −0.56) = 1 − Φ(0.56) = 1 − 0.7123 = 0.2877 ⇒ 28.8%.
- 3. Step 1: P(W < 8.4) = 0.7509 ⇒ z₁ = 0.677. Eq 1: 8.4 − μ = 0.677σ. Step 2: P(W > 9.2) = 0.1385 ⇒ P(W < 9.2) = 0.8615 ⇒ z₂ = 1.087. Eq 2: 9.2 − μ = 1.087σ. Step 3: Subtract: 0.8 = 0.410σ ⇒ σ = 1.95. Step 4: μ = 9.2 − 1.087(1.951) = 7.08.
🎯 Learning objectives & Book exercises
- Understand when and why a binomial distribution X ~ B(n, p) can be approximated by a normal distribution.
- State and verify the validity conditions: np > 5 and n(1 − p) > 5.
- Calculate the parameters of the approximating normal curve: μ = np and σ² = np(1 − p).
- Understand the necessity of the continuity correction (±0.5) when passing from a discrete integer variable to a continuous model.
- Book exercises covered: Exercise 8E and matching review questions.
8.4 The normal approximation to the binomial distribution
- Why approximate? When n is large (e.g. n = 100, 400, or 888), calculating cumulative binomial probabilities by adding individual terms pʳ (1 − p)ⁿ⁻ʳ is exceptionally tedious and error-prone.
- Convergence to the bell curve: as n increases, the discrete binomial probability histogram becomes increasingly symmetrical and smooth, closely hugging a Gaussian normal distribution.
- Necessary conditions:
np > 5 and n(1 − p) > 5 (or nq > 5)
Always check and state both conditions explicitly in exam solutions! - Parameters of the approximating continuous model Y:
- Mean: μ = E(X) = np
- Variance: σ² = Var(X) = np(1 − p) = npq
- Standard deviation: σ = √(npq)
- Hence: X ~ B(n, p) ≈ Y ~ N(np, np(1 − p))
⭐ Key Point 8.4: The Continuity Correction Rules
In the discrete binomial distribution, integer k is represented by a vertical column of width 1, extending from k − 0.5 to k + 0.5. When replacing bars with area under a continuous curve Y, shift boundaries by ±0.5 according to whether the endpoint integer is included or excluded:
| Discrete Binomial Event | Discrete Integers Included | Continuous Normal Region (Y) | Reasoning / Boundary Shift |
|---|---|---|---|
| P(X = k) | Exactly {k} | P(k − 0.5 ≤ Y ≤ k + 0.5) | The entire bar of width 1 centered on k |
| P(X ≥ k) [at least k] | {k, k+1, k+2, ...} | P(Y ≥ k − 0.5) | Includes bar k ⇒ starts at left edge of k |
| P(X > k) [more than k] | {k+1, k+2, ...} | P(Y ≥ k + 0.5) | Excludes bar k ⇒ starts at right edge of k |
| P(X ≤ k) [at most k] | {..., k-2, k-1, k} | P(Y ≤ k + 0.5) | Includes bar k ⇒ extends to right edge of k |
| P(X < k) [fewer than k] | {..., k-2, k-1} | P(Y ≤ k − 0.5) | Excludes bar k ⇒ stops at left edge of k |
| P(a ≤ X ≤ b) | {a, a+1, ..., b} | P(a − 0.5 ≤ Y ≤ b + 0.5) | Includes both outer edges |
| P(a < X < b) | {a+1, ..., b-1} | P(a + 0.5 ≤ Y ≤ b − 0.5) | Excludes bars a and b |
🎲 Interactive Tool: Binomial Approximation & Continuity Correction Solver
Enter the binomial parameters n and p, pick your target probability statement, and calculate the approximating normal probability with full step-by-step working.
💡 Worked Example 8.14: Coin Tossing & Continuity Correction
A fair coin is tossed 888 times. Find, by use of a suitable approximation, the probability of obtaining at least 450 heads.
- Step 1: Define discrete model and test validity conditions
Let X be the number of heads obtained: X ~ B(888, 0.5).
Check conditions:
• np = 888 × 0.5 = 444 > 5
• n(1 − p) = 888 × 0.5 = 444 > 5.
Both conditions are satisfied; a normal approximation is fully justified. - Step 2: Calculate mean and variance of continuous approximating model Y
Mean: μ = np = 444.
Variance: σ² = np(1 − p) = 888 × 0.5 × 0.5 = 222.
Standard deviation: σ = √222 ≈ 14.8997.
So Y ~ N(444, 222). - Step 3: Apply continuity correction to discrete event X ≥ 450
The event "at least 450" includes 450. On the continuous scale, the bar for 450 starts at 450 − 0.5 = 449.5.
Therefore: P(X ≥ 450) ≈ P(Y ≥ 449.5). - Step 4: Standardise continuous boundary y = 449.5
z = = ≈ 0.369. - Step 5: Look up standard normal probability and conclude
P(Y ≥ 449.5) = P(Z ≥ 0.369) = 1 − Φ(0.369).
From tables: Φ(0.36) = 0.6406; ADD 9 gives +0.0033 ⇒ Φ(0.369) = 0.6439.
1 − 0.6439 = 0.3561.
Answer: 0.356 (to 3 s.f.).
💡 Worked Example 8.12: Approximating an Inclusive Interval
The discrete random variable X ~ B(100, 0.4). Use a suitable approximation to find P(35 ≤ X ≤ 45).
- Step 1: Check approximation validity
np = 100(0.4) = 40 > 5, and nq = 100(0.6) = 60 > 5. Normal approximation is justified. - Step 2: Find continuous normal parameters
Mean μ = 40, variance σ² = 100(0.4)(0.6) = 24 ⇒ σ = √24 ≈ 4.8990.
Approximating model: Y ~ N(40, 24). - Step 3: Apply continuity corrections at both boundaries
Since both 35 and 45 are included, the lower boundary expands left to 35 − 0.5 = 34.5, and the upper boundary expands right to 45 + 0.5 = 45.5.
P(35 ≤ X ≤ 45) ≈ P(34.5 ≤ Y ≤ 45.5). - Step 4: Standardise both boundaries
z₁ = (34.5 − 40)/√24 = −5.5 / 4.8990 = −1.123.
z₂ = (45.5 − 40)/√24 = +5.5 / 4.8990 = +1.123. - Step 5: Compute central interval probability
P(−1.123 ≤ Z ≤ 1.123) = 2Φ(1.123) − 1.
From tables: Φ(1.123) = 0.8686 + 0.0006 = 0.8692.
Compute: 2(0.8692) − 1 = 1.7384 − 1 = 0.7384.
Answer: 0.738 (or 0.739).
✏️ Try it: Normal approximation to binomial
- For X ~ B(50, 0.6), state the approximating normal distribution and write the probability statement for P(X > 26) with continuity correction.
- A multiple-choice exam has 120 questions, each with 4 options. A student guesses every answer at random. Use a normal approximation to find the probability of getting at least 35 correct.
- Decide whether B(40, 0.13) can be well-approximated by a normal distribution, giving reasons.
Answers
- 1. Step 1: np = 50(0.6) = 30 > 5, nq = 20 > 5. Mean μ = 30, σ² = 50(0.6)(0.4) = 12. Model: Y ~ N(30, 12). Step 2: X > 26 means X ≥ 27, so the continuity correction gives P(Y ≥ 26.5).
- 2. Step 1: n = 120, p = 0.25. np = 30 > 5, nq = 90 > 5. μ = 30, σ² = 120(0.25)(0.75) = 22.5 ⇒ σ = √22.5 ≈ 4.7434. Step 2: P(X ≥ 35) becomes P(Y ≥ 34.5). Step 3: z = (34.5 − 30)/4.7434 = 4.5/4.7434 = 0.949. Step 4: 1 − Φ(0.949) = 1 − 0.8287 = 0.171.
- 3. Step 1: n = 40, p = 0.13. np = 40 × 0.13 = 5.2 > 5; nq = 40 × 0.87 = 34.8 > 5. Step 2: While technically np = 5.2 marginally exceeds 5, it is very close to 5, meaning the approximation will be relatively poor. Answer: Yes, it can be approximated (np = 5.2 > 5, nq = 34.8 > 5), with μ = 5.2 and σ² = 4.524, but it is borderline accurate.
8.5 Chapter summary & checklist
- Continuous random variables: modelled by a probability density function f(x) ≥ 0 where total area = 1. Probability corresponds to area under the curve; P(X = c) = 0 for any single point c.
- Normal distribution X ~ N(μ, σ²): symmetrical bell curve centered at μ, with points of inflection at μ ± σ. Mean = median = mode = μ.
- Standard normal variable Z ~ N(0, 1): the universal benchmark distribution with mean 0 and standard deviation 1. Cumulative distribution function denoted by Φ(z) = P(Z < z).
- Reflection symmetry: Φ(−z) = 1 − Φ(z). For upper tails: P(Z > z) = 1 − Φ(z). For central intervals: P(−z < Z < z) = 2Φ(z) − 1.
- Standardisation: converts any X ~ N(μ, σ²) to Z ~ N(0, 1) using z = . In reverse: x = μ + zσ.
- Unknown parameters μ and σ: two independent probability conditions translate into two simultaneous linear equations: x₁ − μ = z₁σ and x₂ − μ = z₂σ, solved by elimination.
- Normal approximation to binomial B(n, p): valid when np > 5 and n(1 − p) > 5. Approximating normal has mean μ = np and variance σ² = np(1 − p).
- Continuity correction (±0.5): required because discrete integer bars of width 1 are approximated by continuous area. Always check endpoints (e.g. X ≥ k → Y ≥ k − 0.5).
📋 Printable assignment & Cross-topic review
A complete 47-page printable assignment is ready for this chapter, containing 105 practice questions (213 marks) covering Exercises 8A to 8E, End-of-chapter review 8, Cross-topic review exercise 3 (covering chapters 6 to 8 and the complete S1 paper), and original Check yourself questions.
Chapter 8 Interactive Quiz
Source: Cambridge International AS & A Level Mathematics: Probability & Statistics 1 Coursebook (Cambridge University Press, Dean Chalmers), Chapter 8 (The normal distribution) and Cross-topic review exercise 3. Explanations, worked examples, diagrams, the interactive simulation, the interactive tools, the Try-it answers and the quiz are Guru Jazzy's own. This page is not made or endorsed by Cambridge University Press or Cambridge Assessment.