Start from the area to the left
For Z ~ N(0, 1), a table gives , the area to the left of z. Every other area is built from these.
At z = 0 the area to the left is a half, since the curve is symmetric about 0: P(Z < 0) = 0.5. Further right, the table gives P(Z < 1) = 0.8413.
The area to the left of z = 0 is half of the whole: P(Z < 0) = 0.5.
A right-hand tail
The area to the right of z is everything that is not to its left. The whole area is 1, so . For z = 1: P(Z > 1) = 1 − 0.8413 = 0.1587.
Whether the end point is included makes no difference, since a single value has probability 0: is also 0.1587. By symmetry, P(Z > 1) is the same as P(Z < −1).
The tail to the right of z = 1. The table gives the 0.8413 to its left, so the tail is 1 − 0.8413 = 0.1587.
A band between two values
The area between a and b is everything to the left of b, take away everything to the left of a: . Subtract the smaller area from the larger.
For example, .
A band with a negative edge needs symmetry first. For P(−1.5 < Z < 0.5), the table gives and , so . Then P(−1.5 < Z < 0.5) = 0.6915 − 0.0668 = 0.6247.
On the original scale
The lifetimes L of a brand of battery follow , in hours. What share of the batteries last more than 560 hours? Standardize first: . Then P(L > 560) = P(Z > 1.5) = 1 − 0.9332 = 0.0668, so about 6.7% of the batteries.
What share last between 440 and 520 hours? The two edges standardize to and , so the answer is P(−1.5 < Z < 0.5) = 0.6247, about 62%.
A sketch is a quick check. The band from 440 to 520 holds the peak, so it should hold well over half of the area, and 0.6247 does. A tail that starts 1.5 standard deviations out should be small, and 0.0668 is.
The battery lifetimes, , with the band from 440 to 520 hours shaded. Its edges are 1.5 standard deviations below the mean and half a standard deviation above it, and it holds 0.6247 of the area.
The usual mistakes
Giving the area to the left as the answer for a tail. The table gives 0.8413 for z = 1, and that is P(Z < 1); the tail P(Z > 1) is 1 − 0.8413.
Running a band from the far left. P(0 < Z < 1.5) is not 0.9332, which includes the half below 0; it is 0.9332 − 0.5 = 0.4332.
Subtracting the wrong way round. A probability is never negative, so with a below b has the two areas in the wrong order.
Worked example: Heights of Adult Women Modeled as Normal, a Proportion Between Two Heights and the Number Above One
Question The heights of adult women in a town are modeled by H ∼ N(165, 62), in centimeters. (a) Find the proportion of women whose heights lie between 159 cm and 174 cm. (b) A clothing shop stocks a tall range for women over 177 cm. Out of 500 women, how many would you expect to be over 177 cm?
1.Standardize each height with z = h − 1656: 159 cm gives z = 159 − 1656 = −1, and 174 cm gives z = 174 − 1656 = 1.5.
Under each height is its z-score, the number of standard deviations from the mean: 159 cm is at z = −1 and 174 cm at z = 1.5. 2.Read the table: Φ(1.5) = 0.9332. By symmetry, Φ(−1) = 1 − Φ(1) = 1 − 0.8413 = 0.1587.
The shaded tail left of z = −1 is Φ(−1) = 1 − Φ(1) = 0.1587, by the symmetry of the curve. 3.(a) P(159 < H < 174) = Φ(1.5) − Φ(−1) = 0.9332 − 0.1587 = 0.7745, so about 77% of the women.
(a) The area between the two heights is Φ(1.5) − Φ(−1) = 0.9332 − 0.1587 = 0.7745. 4.For 177 cm, z = 177 − 1656 = 2, and P(H > 177) = 1 − Φ(2) = 1 − 0.9772 = 0.0228.
Above 177 cm is the tail beyond z = 2: 1 − Φ(2) = 0.0228. 5.(b) The expected number out of 500 is 500 × 0.0228 = 11.4, so about 11 women. Check: 177 cm is 2 standard deviations above the mean, and a little over 2% of a normal distribution lies that far above its mean.
(b) Out of 500 women, 500 × 0.0228 = 11.4, so about 11 are expected to be over 177 cm.
Answer: (a) 0.7745; (b) 11.4, so about 11 women
Common mistakes
- Dividing by the variance: 174 − 16536 = 0.25. N(165, 62) gives the variance 36; the z-score divides by the standard deviation, 6.
- Reading Φ(−1) as 0.8413. The table holds positive z only, and 0.8413 is the area to the left of +1; the area to the left of −1 is 1 − 0.8413 = 0.1587.
More probability distributions problems, worked step by step →