Source
Adapted from my own handwritten and hand-typed notes from Year 1 to Year 3 of my EEE degree
For large the binomial probabilities are tedious to sum, so we approximate the binomial by a normal curve with the same mean and variance.
Condition and parameters
If with and , then approximately
Continuity correction
The binomial is discrete (mass only at integers) while the normal is continuous, so each integer is replaced by the interval . The lectures call this the envelope principle. Here is binomial and is normal.
| Binomial | Normal approximation |
|---|---|
Tip: first rewrite strict inequalities with or on integers, then widen by 0.5.
Worked example
A fair coin is tossed 100 times. = number of tails, . Since , use .
Solution:
- exactly 60:
- 48 to 53 inclusive:
- fewer than 45:
The exact binomial value for is also 0.0108, so the approximation is very good here. The standardising step is the same as in Normal Distribution.