de Moivre-Laplace Theorem — Definition, Formula & Example
The de Moivre-Laplace theorem is the result that a binomial count, such as the number of heads in many coin flips, follows a bell-shaped normal curve once the number of trials is large. It lets you replace a long sum of binomial probabilities with one area under the normal curve.
Let with fixed , and let . Then the standardized variable converges in distribution to the standard normal . Equivalently, for all real , as . It is the special case of the Central Limit Theorem for sums of independent Bernoulli trials. De Moivre proved it for in 1733, and Laplace extended it to general in 1812.
Key Formula
Where:
- = Number of successes in n independent trials
- = Number of trials
- = Probability of success on each trial, with 0 < p < 1
- = Mean of the binomial distribution
- = Standard deviation of the binomial distribution
- = Cumulative distribution function of the standard normal distribution
Worked Example
Problem: You guess randomly on a 100-question true/false quiz, so each answer is correct with probability 0.5. Use the de Moivre-Laplace theorem to estimate the probability that you get 60 or more correct.
Step 1: The number correct is . Find its mean and standard deviation.
Step 2: Apply the continuity correction. The event covers the bar from 59.5 to 60.5 and everything above it, so start the normal area at 59.5. Then standardize.
Step 3: Find the upper-tail area using a standard normal table, where .
Answer: The probability is about 0.0287, or roughly 2.9%. The exact binomial value is about 0.0284, so the approximation is off by only about 0.0003.
Why It Matters
The one-proportion z-test and the confidence interval for a proportion in AP Statistics both rest on this theorem. It is also the reason for the textbook check that and before you use a normal model. Some books use 5 instead of 10. Historically, it was the first version of the Central Limit Theorem ever proved.
Common Mistakes
Mistake: Skipping the continuity correction and computing , which gives .
Correction: A binomial count is discrete, while the normal curve is continuous. Shift each boundary by 0.5 so the normal area covers the whole bar for each included value. Here, use 59.5. The corrected estimate 0.0287 is far closer to the exact 0.0284.
