Normal Approximation Calculator

Q: What is the difference between the binomial, Poisson, and normal distributions?

The binomial distribution models the number of successes in n independent trials each with probability p of success. It requires fixed n and independent, identically distributed Bernoulli trials. The Poisson distribution models the number of events in a fixed interval when events occur independently at a constant average rate λ - it has no upper bound. The normal distribution is continuous and bell-shaped; it arises as the limit of both binomial (large n) and Poisson (large λ) distributions, making it a universal approximation tool.

Q: Why is the normal approximation used if exact calculations exist?

For large n (e.g. n = 1000, p = 0.3), computing exact binomial probabilities requires evaluating combinations like C(1000,300), which involves astronomically large factorials. Before computers, the normal approximation was the only practical way to compute such probabilities. Even today, it provides intuition (via Z-scores) and is used in statistical inference - for example, the normal approximation underlies the z-test for proportions and the construction of confidence intervals for proportions.

Q: What is a Z-score in the context of normal approximation?

After applying continuity correction, the adjusted k value (k ± 0.5) is converted to a Z-score: Z = (k ± 0.5 − μ) / σ. This Z-score represents how many standard deviations from the mean the value falls in the approximating normal distribution. The probability is then read from the standard normal CDF. For example, if μ=50, σ=5, and we want P(X ≤ 55), we compute Z = (55.5 − 50) / 5 = 1.10, giving P ≈ 0.864.

Q: How accurate is the normal approximation compared to the exact binomial?

When the validity conditions are met (np ≥ 10, nq ≥ 10), the normal approximation typically gives probabilities within 0.01–0.02 of the exact binomial. The accuracy improves rapidly with n. With continuity correction, the approximation is substantially more accurate than without it, particularly for probabilities involving specific values of X (P(X = k) type). For the highest accuracy with any n and p, use exact binomial calculations.

Q: Can I use normal approximation for a range of values, P(a ≤ X ≤ b)?

Yes. For P(a ≤ X ≤ b) with continuity correction, approximate as P(a − 0.5 ≤ X ≤ b + 0.5) under the normal distribution: this equals Φ((b + 0.5 − μ)/σ) − Φ((a − 0.5 − μ)/σ). This calculator handles this case in the Range option. Make sure that the lower bound is less than the upper bound.

Approximate binomial or Poisson probabilities using the normal distribution with continuity correction.

📖 What is Normal Approximation?

The normal approximation is a technique for estimating probabilities from discrete distributions - most commonly the binomial and Poisson distributions - using the continuous normal (Gaussian) distribution. It is justified by the Central Limit Theorem, which states that the sum of a large number of independent random variables approaches a normal distribution, regardless of the underlying distribution.

For the binomial distribution B(n, p), when n is large enough that both np ≥ 10 and n(1−p) ≥ 10, the distribution is approximately normal with mean μ = np and standard deviation σ = √(np(1−p)). For the Poisson distribution Pois(λ), when λ ≥ 10, it is approximately normal with μ = λ and σ = √λ.

Because binomial and Poisson are discrete and the normal is continuous, a continuity correction is applied. To approximate P(X ≤ k), we use P(X ≤ k + 0.5) under the normal. To approximate P(X = k), we use P(k − 0.5 ≤ X ≤ k + 0.5). This half-unit shift dramatically improves accuracy and should always be applied.

Normal approximation is widely used in hypothesis testing for proportions, quality control sampling, insurance risk modelling, and election polling. Understanding when and how to apply it - and its limitations - is fundamental to applied statistics.

📐 Formulas

Binomial: μ = np, σ = √(np(1−p))

Binomial to Normal: X ~ B(n, p) ≈ N(np, np(1−p)) when np ≥ 10 and n(1−p) ≥ 10.

Poisson to Normal: X ~ Pois(λ) ≈ N(λ, λ) when λ ≥ 10.

Continuity correction:

P(X ≤ k) ≈ Φ((k + 0.5 − μ) / σ)

P(X ≥ k) ≈ 1 − Φ((k − 0.5 − μ) / σ)

P(X = k) ≈ Φ((k + 0.5 − μ) / σ) − Φ((k − 0.5 − μ) / σ)

P(a ≤ X ≤ b) ≈ Φ((b + 0.5 − μ) / σ) − Φ((a − 0.5 − μ) / σ)

Validity conditions: Binomial: np ≥ 10 AND n(1−p) ≥ 10. Poisson: λ ≥ 10.

All variables: n = number of trials; p = probability of success; q = 1 − p; k = specific value; λ = Poisson rate; μ = mean; σ = standard deviation; Φ = standard normal CDF.

📖 How to Use This Calculator

Select Binomial or Poisson. For Binomial, enter the number of trials n and the probability p. For Poisson, enter the rate parameter λ.

Choose the probability type (P(X≤k), P(X≥k), P(X=k), or a range), then enter the value k (or bounds a and b for a range).

Click Calculate Probability. See whether the normal approximation is valid, the mean and SD, Z-score with continuity correction, and the approximated probability.

💡 Example Calculations

Example 1 - Quality control: defective items

A factory produces 200 items. Each has a 10% chance of being defective. Find P(X ≤ 25) using normal approximation.

μ = 200 × 0.10 = 20. σ = √(200 × 0.10 × 0.90) = √18 = 4.243. np = 20 ≥ 10 and nq = 180 ≥ 10 - approximation valid.

Z = (25.5 − 20) / 4.243 = 1.296. P(X ≤ 25) ≈ Φ(1.296) ≈ 0.9026. There is about a 90% chance of 25 or fewer defectives.

Result = P(X ≤ 25) ≈ 0.9026 (90.3%)

Try this example →

Example 2 - Election poll: estimating support

In a poll of n = 500 voters, p = 0.52 support candidate A. Find P(X ≥ 270) to assess whether a majority is likely.

μ = 500 × 0.52 = 260. σ = √(500 × 0.52 × 0.48) = √124.8 = 11.17. Both np = 260 ≥ 10 and nq = 240 ≥ 10.

Z = (269.5 − 260) / 11.17 = 0.850. P(X ≥ 270) ≈ 1 − Φ(0.850) ≈ 0.198. About a 20% chance of 270 or more in the sample - the expected 260 is well below 270.

Result = P(X ≥ 270) ≈ 0.198 (19.8%)

Try this example →

Example 3 - Insurance: Poisson claims

An insurer expects λ = 50 claims per month. Find P(X ≤ 60) to assess reserve adequacy.

μ = λ = 50. σ = √50 = 7.071. λ = 50 ≥ 10 - approximation valid.

Z = (60.5 − 50) / 7.071 = 1.485. P(X ≤ 60) ≈ Φ(1.485) ≈ 0.931. Reserves covering 60 claims are sufficient in about 93% of months.

Result = P(X ≤ 60) ≈ 0.931 (93.1%)

Try this example →

❓ Frequently Asked Questions

When is the normal approximation to the binomial valid?+

The rule of thumb most commonly used is that both np ≥ 10 and n(1−p) ≥ 10 should hold. Some textbooks use np ≥ 5 and nq ≥ 5 as a looser criterion. When p is close to 0.5, the binomial distribution is nearly symmetric and the normal approximation works well even for moderate n. When p is very close to 0 or 1, the distribution is highly skewed and the approximation requires very large n. This calculator checks both conditions and warns if they are not met.

What is continuity correction and why is it needed?+

The binomial and Poisson distributions are discrete - X can only take integer values. The normal distribution is continuous. When approximating P(X = k) with a continuous distribution, we approximate it as P(k − 0.5 ≤ X ≤ k + 0.5). For cumulative probabilities: P(X ≤ k) is approximated as P(X ≤ k + 0.5) and P(X ≥ k) as P(X ≥ k − 0.5). Without continuity correction, the approximation can be off by half a probability unit, which matters especially for small probabilities.

How does the Poisson to normal approximation work?+

For a Poisson distribution with parameter λ (the mean and variance), when λ is large, the distribution becomes approximately normal with mean μ = λ and standard deviation σ = √λ. The approximation improves as λ increases - the conventional rule is λ ≥ 10. For λ < 10, use the exact Poisson formula. The same continuity correction applies: for P(X ≤ k), use the normal CDF evaluated at k + 0.5.

What is the difference between the binomial, Poisson, and normal distributions?+

The binomial distribution models the number of successes in n independent trials each with probability p of success. It requires fixed n and independent, identically distributed Bernoulli trials. The Poisson distribution models the number of events in a fixed interval when events occur independently at a constant average rate λ - it has no upper bound. The normal distribution is continuous and bell-shaped; it arises as the limit of both binomial (large n) and Poisson (large λ) distributions, making it a universal approximation tool.

Why is the normal approximation used if exact calculations exist?+

For large n (e.g. n = 1000, p = 0.3), computing exact binomial probabilities requires evaluating combinations like C(1000,300), which involves astronomically large factorials. Before computers, the normal approximation was the only practical way to compute such probabilities. Even today, it provides intuition (via Z-scores) and is used in statistical inference - for example, the normal approximation underlies the z-test for proportions and the construction of confidence intervals for proportions.

What is a Z-score in the context of normal approximation?+

After applying continuity correction, the adjusted k value (k ± 0.5) is converted to a Z-score: Z = (k ± 0.5 − μ) / σ. This Z-score represents how many standard deviations from the mean the value falls in the approximating normal distribution. The probability is then read from the standard normal CDF. For example, if μ=50, σ=5, and we want P(X ≤ 55), we compute Z = (55.5 − 50) / 5 = 1.10, giving P ≈ 0.864.

How accurate is the normal approximation compared to the exact binomial?+

When the validity conditions are met (np ≥ 10, nq ≥ 10), the normal approximation typically gives probabilities within 0.01–0.02 of the exact binomial. The accuracy improves rapidly with n. With continuity correction, the approximation is substantially more accurate than without it, particularly for probabilities involving specific values of X (P(X = k) type). For the highest accuracy with any n and p, use exact binomial calculations.

Can I use normal approximation for a range of values, P(a ≤ X ≤ b)?+

Yes. For P(a ≤ X ≤ b) with continuity correction, approximate as P(a − 0.5 ≤ X ≤ b + 0.5) under the normal distribution: this equals Φ((b + 0.5 − μ)/σ) − Φ((a − 0.5 − μ)/σ). This calculator handles this case in the Range option. Make sure that the lower bound is less than the upper bound.

Normal Approximation Calculator

📖 What is Normal Approximation?

📐 Formulas

📖 How to Use This Calculator

💡 Example Calculations

Example 1 - Quality control: defective items

Example 2 - Election poll: estimating support

Example 3 - Insurance: Poisson claims

❓ Frequently Asked Questions

🔗 Related Calculators

📌 Quick Tips