A binomial distribution counts the number of successes in a fixed number of independent trials, such as correct answers, defective items, or heads in coin flips. When the number of trials is large, calculating exact binomial probabilities can become slow and repetitive. The normal approximation replaces the binomial bars with a smooth bell curve that is much easier to work with.
This matters because it lets students estimate binomial probabilities using z-scores and normal tables or calculators.
Understanding Statistics: Normal Approximation to the Binomial
The approximation works because a total is built from many small random contributions. In each trial, the count either stays where it is or increases by one. Across a large set of trials, unusually high and unusually low totals become less common than totals near the centre.
The combined pattern often becomes smooth and mound shaped. This is connected to the central limit theorem, a major idea in statistics. It explains why normal curves appear in measurements, sampling results, manufacturing data, and many totals made from separate random events.
A binomial count can take only whole-number values. There can be 42 successes, but not 42.3 successes. A normal variable is different because it can take every value along a continuous line.
The continuity correction links these two views. The bar for a count of 42 covers the stretch from 41.5 to 42.5 on the continuous scale. Therefore, an area under the curve should usually begin or end halfway between two possible counts.
This small adjustment matters most when the count is not extremely large. Without it, an estimate may systematically include too much or too little area.
The quality of the approximation depends on the shape of the original distribution. A binomial distribution is fairly balanced when success and failure have similar chances. It becomes lopsided when one outcome is rare.
For example, if defects occur only three percent of the time, most production batches have few defects and the distribution has a long tail toward larger counts. A symmetric bell curve cannot copy that shape well, especially near zero.
Students should check that the expected numbers of successes and failures are both comfortably large. They should be particularly careful with probabilities in far tails, where even a small shape mismatch can change the result noticeably.
In practice, begin by deciding whether the situation really has independent trials with the same success chance each time. Repeated coin flips fit this model well. Drawing cards without replacement does not, because each draw changes the next one.
Next, find the centre and spread of the count, then place the requested whole-number event on the continuous scale using the half-unit boundaries. Convert that boundary to a z-score and use the normal curve area. In quality control, this can estimate whether a shipment has an unusually large number of faulty items.
In polling, it helps describe how much a count of respondents can vary by chance. The final answer is an estimate, so it should be reported with sensible rounding and interpreted as approximate rather than exact.
Key Facts
- For X ~ Bin(n, p), the mean is μ = np.
- For X ~ Bin(n, p), the standard deviation is σ = sqrt(npq), where q = 1 - p.
- Use the normal approximation when np ≥ 10 and nq ≥ 10.
- Approximate X ~ Bin(n, p) with Y ~ N(np, npq).
- Convert values using z = (x - μ) / σ after applying any needed continuity correction.
- Continuity correction: P(X ≤ k) ≈ P(Y < k + 0.5) and P(X ≥ k) ≈ P(Y > k - 0.5).
Vocabulary
- Binomial distribution
- A probability distribution for the number of successes in n independent trials when each trial has the same success probability p.
- Normal approximation
- A method that estimates a binomial probability using a normal distribution with the same mean and standard deviation.
- Continuity correction
- An adjustment of 0.5 added or subtracted when using a continuous normal curve to estimate a discrete binomial probability.
- Success probability
- The value p, which is the probability that one binomial trial results in a success.
- Z-score
- A standardized value that tells how many standard deviations a data value is from the mean.
Common Mistakes to Avoid
- Skipping the condition check. The normal approximation can be inaccurate if np or nq is too small because the binomial distribution may be too skewed.
- Forgetting the continuity correction. A binomial variable uses whole-number counts, so the normal boundary should be shifted by 0.5 to better match the bar area.
- Using σ = npq instead of σ = sqrt(npq). The variance is npq, but the standard deviation used in z = (x - μ) / σ is the square root of the variance.
- Treating p and q as the same value. Since q = 1 - p, using the wrong q gives the wrong spread and changes the final probability.
Practice Questions
- 1 A multiple-choice test has 80 questions, and a student guesses with probability p = 0.25 of getting each correct. Use the normal approximation with continuity correction to estimate P(X ≥ 25).
- 2 A factory produces items with a defect rate of 0.08. In a sample of 200 items, use the normal approximation with continuity correction to estimate P(X ≤ 10 defects).
- 3 Explain why the continuity correction improves the normal approximation when estimating a binomial probability, and describe when the normal approximation should not be used.