JC2 · H2 Mathematics · Notes
Translate totals, weighted observations, absolute differences, means and counts into the correct normal model, with complete probability calculations.
Normal Distribution summary and formula sheet: every method on 5 pages
Mr Ian Ang’s summary of the whole normal distribution chapter on five pages.
Page 1 of 5 The normal curve Part 1: one variable
H2 Math summary notes
Normal Distribution formulas and methods, section by section
Page 1 of 5The normal curvePart 1: one variable
X ~ N(μ, σ²) and the normal curve: area, symmetry and the 68-95-99.7 rule
- Write the variance in the bracket: , with (mean) and (variance). On the GC key . The area under the curve between and is .
- (1) The total area under the curve is 1. (2) (a single point), so . Warning: binomial has , but normal has .
- (3) The curve is symmetric about and gets closer to the -axis as . (4) A smaller gives a taller and narrower curve (more concentrated about ); the same with a different gives the same shape, translated.
- The bands from to are 2.35%, 13.5%, 34%, 34%, 13.5%, 2.35%. Within of : 68%; : 95%; : 99.7%. (GC: 0.341).
- Quartiles: , and .
When a normal model is not suitable: check the undesired range
- Example: the waiting time minutes has mean 14 and standard deviation 8. Explain why a normal model may not be suitable.
- Assume . Then .
- is non-negligible, and time cannot be negative, so is not appropriate.
- Warning: check the probability of the undesired range (a time, a mass: must be ).
Sketching normal curves, comparing them
- (a) : sketch the distribution of for . Label the mean (dashed) at 40 and the two ends of the given range, 25 and 55, and keep it symmetric.
- (b) On one diagram, sketch and . The smaller is taller and narrower (both areas are 1).
- (c) . Without calculating, explain with a diagram how and compare. Both intervals are 3 units wide and is nearer the mean, so has the larger area: .
- Warning: 'with the aid of a diagram': the diagram gives the reason.
Page 2 of 5Probabilities and inverse normal on the GC
Probabilities on the GC: normalcdf
| We want | Interval | You key in | You write |
|---|---|---|---|
| normalcdf(27, E99, 30, ) | (3 s.f.) | ||
| normalcdf(E99, 34, 30, ) | (3 s.f.) | ||
| normalcdf(27, 34, 30, ) | (3 s.f.) |
- (1) Let be the random variable denoting ... (2) Write , with and . On the GC, , not 20.
- Keys: 2nd VARS (DISTR) 2: normalcdf( lower, upper, , , then Paste.
- E99 means , a large number: upper limit E99 for 'more than', lower limit E99 for 'less than'. Write 'normalcdf' under the answer.
Inverse normal: invNorm with LEFT, RIGHT, CENTER
| Left area | Right area | Centre area | |
|---|---|---|---|
| we want | |||
| key in, and get | invNorm(0.3, 0, 1, LEFT) gives | invNorm(0.3, 0, 1, RIGHT) gives | invNorm(0.3, 0, 1, CENTER) gives , so |
| an inequality | ; left area: keep the inequality | ; right area: flip the inequality |
- normalcdf: (area asked); invNorm: (value asked). Keys: 2nd VARS 3: invNorm( area, , , Tail (LEFT, CENTER, RIGHT).
- With area 0.3 on the standard normal: left area gives ; right area gives ; centre area gives , so .
- For an inequality: (left area: keep the inequality); (right area: flip the inequality); .
- Example: . Find the mass exceeded by 70% of the items. : invNorm(0.7, 40, 3, RIGHT) gives (3 s.f.). 'Exceeded' means , RIGHT tail.
- Warning: CENTER is the area symmetric about ; it returns two values.
Page 3 of 5Standardising and symmetry
Standardising to find μ or σ: standardise, then invNorm
- Use when or is unknown. Step 1: standardise. Step 2: invNorm (with : 0, : 1; the current variable is ).
- One unknown: and ; find . The probability stays the same. Standardise: . invNorm with area 0.1, RIGHT, gives , so (3 s.f.).
- When is not a whole number, use the GC graph (check): with , and . The intersection is , so . A whole number unknown (, ) uses the GC table (section 10).
- Two unknowns: , and ; find and . From : , so (1). From : , so (2).
- Using the GC to solve (1) and (2): and (3 s.f.).
- Warning: standardising keeps the inequality sign. For or a total, standardise with its own mean and variance (section 10).
Symmetry arguments: equal tails, equidistant values and within k of the mean
- (a) ; find . Equal tails put at the midpoint: .
- (b) , , . Show that . Standardise: , so . Hence , so .
- (c) and 70% of lies within 4 of the mean; find . . Dividing by , . Open up the inequality: . invNorm, CENTER, gives , so (3 s.f.).
- (d) is already there: divide by only. (3 s.f.).
- Warning: for 'within of the mean' and 'differs from its mean by less than ', insert the modulus.
Page 4 of 5Linear combinations and 'differs by'Part 2: several variables
Linear combinations: E and Var line by line
- , independent: and : the variances always add. In general ; the coefficient is squared.
- (subscript: 3 different items) but (coefficient: 1 item, 3 times). They are NOT interchangeable.
- Reduce by 15%: multiply by 0.85 (0.85 at Var); increase by 6%: multiply by 1.06. Cost is price per unit mass times mass: \\dfrac{4.5}{1000}\mathrm{N}\!\left(\dfrac{45}{60},\left(\dfrac{6}{60}\right)^{2}\right)5+0.8t$.
- Example: melons have mass g and sell at \P\sim \mathrm{N}(900,60^{2})T=\dfrac{4.5}{1000}(M_1+M_2+M_3)+\dfrac{3}{1000}(P_1+P_2)3M$).
- and . So (on the GC, ) and (3 s.f.) by normalcdf.
- ' more than ': , with the left side the variable and the right side a constant. Assume all the masses are independent of one another. Wrong: 'the mass of a melon is independent of the mass of a pineapple', which misses , , among themselves.
"Differs by": the modulus
- The difference between and is ; 'differs' and 'within' mean the modulus of the difference. With : differs by more than 5 is , the first being ; by less than 5 is .
- Example: and . Find the probability that differs from by more than 5. with . So the probability is (3 s.f.).
- Two of the same item: . With and , .
- is less than by at most 2: (1) and (2) , so . Do not miss out 0. By at least 2: (1) and (2) , so .
- Warning: 'differs by' has two tails; 'less than by' fixes the sign.
Page 5 of 5The sample mean and a normal probability inside a count
The sample mean X̄: N(μ, σ²/n) for the same variable
| 15 | |
|---|---|
| 16 | |
| 17 |
- is the average of over objects; only the variance is affected. (no squaring at Var), and (coefficient: square at Var). Warning: is for the same variable only; different items make a combination, as in (b).
- (a) Pears have mass g. Find the probability that the mean mass of 9 pears is less than 146 g. (GC: ), and (3 s.f.).
- (b) Apples have mass g. Find the probability that the mean mass of 2 pears and 3 apples is more than 135 g. , and (no squaring at 2 and 3, coefficient squared at ). So (3 s.f.).
- (c) Find the least for which the mean mass of pears has . , keyed as with . The table gives : 0.053292 ; : 0.047790 ; : 0.042902 . The least is 16.
A normal probability inside a count: find p from the normal, then B(n, p)
- g (cookies); .
- (a) 8 cookies are chosen. Find . Trial: a cookie; success: mass g. Let be the number of cookies, out of 8, with mass g: . (3 s.f.). Warning: 'out of trials' means binomial; three ranges (below 28.8, above 32, others) means not binomial, so use a tree.
- (b) 5 cookies: 3 have mass less than 28.8 g and 2 have mass more than 32 g. The tree for has branches , and others, repeated for to . The probability is (3 s.f.); count the number of paths.
- (c) The 6th cookie is the 2nd with mass more than 32 g. Split into to (1 out of 5 above 32) and (the 2nd above 32): (3 s.f.), or .
- (d) The heaviest and lightest of 6 cookies. Heaviest means all 6 are : (3 s.f.). Lightest means all 6 are : (3 s.f.).
Worked examples
Normal Distribution worked examples
Mr Ian’s worked examples for the chapter, with the full working.
Define the quantity whose probability is required
In , the second parameter is the variance. Standardise with the standard deviation , not the variance. For a total, difference, mean, price or percentage change, define that new random variable and find its mean and variance before calculating its probability.
Use a percentile to find an unknown parameter
Recover both mean and standard deviation
Apple masses X grams are normally distributed. Ten percent are below 80 g and fifteen percent exceed 120 g. Find the mean and standard deviation.
Inverse normal calculations use cumulative left areas. Thus the second left area is 0.85, not 0.15:
Multiply by before subtracting the equations:
The mean lies between the two cutoffs and closer to 120, consistent with the unequal tail areas. Keep unrounded values for later parts.
Use symmetry before standardising
For , suppose . Find .
The cutoffs k and lie equal distances on opposite sides of the mean. Reflection swaps left and right tails:
A normal variable has no probability mass at a single point, so strict and non-strict endpoints give the same probability.
Different observations are not multiples of one observation
A combined packet mass
Packet masses are independent. Type A has distribution g and type B has distribution g. Find the probability that two A packets and three B packets have combined mass above 0.55 kg.
Use separate subscripts for separate packets. Convert 0.55 kg to 550 g:
By contrast, means scaling one A observation and one B observation. Its variance is , not 66. The same random mass is being reused in a multiple.
“Differs by” means an absolute difference
Two visits compared with one visit
Times on two visits to shop A each have distribution minutes. A visit to shop B has distribution minutes. Assume all three visit times are mutually independent. Find the probability that the total at A differs from the time at B by at least 15 minutes.
The event is , so add the two disjoint tails. Write q for this probability:
Do not double one tail: D has mean 10, so the two tails cut off by −15 and 15 are unequal. The assumption covers both A visits as well as the B visit. The variable itself is not normal: it cannot take negative values.
Move the “differs by at least” threshold
c = 15 min: left tail 0.0888; right tail 0.3937; total 0.4826.
A sample mean has a smaller variance
Choose enough observations for a one-sided requirement
Independent filling times have distribution seconds. Find the least n such that .
This is exact for any n because the individual times are normal. The required left area is at least 0.90:
The least n is 27. At n = 26 the probability is about 0.8988; at n = 27 it is about 0.9030. The denominator for a mean is , not .
Heaviest and lightest: translate into “all”
The heaviest of eight components
Eight independent component masses each have distribution g. Find the probability the heaviest is below 110 g, then the probability the lightest exceeds 110 g.
The heaviest is below 110 only when every mass is below 110:
Here q is the requested heaviest-mass probability. The lightest exceeds 110 only when every mass exceeds 110:
For “at least one exceeds 110”, use . The maximum and minimum are not themselves normally distributed.
A normal probability becomes a binomial success probability
Count acceptable bolt–hole pairs
A bolt diameter is cm and a hole diameter is cm, independently. A pair is acceptable when the hole is between 0.02 cm and 0.15 cm wider than the bolt. Twenty independent pairs are packed in a box. Find the probability at least 18 are acceptable.
Now change the random variable: V counts acceptable pairs, each with the same success probability p.
There are two independence assumptions: bolt and hole diameters within a pair, and acceptance events across different pairs. Retain the unrounded normal probability when using the binomial model.
Percentage changes and price conversions
Wrapping increases every fruit’s mass by eight percent
Independent fruit masses are g and g. A bag contains three A fruits and two B fruits. Each wrapped fruit weighs 8% more than that fruit alone. The bag itself has negligible mass. Find the probability the bag weighs above 1.7 kg.
It is simpler to convert the event back to T than to round the variance of W:
The wrapper is tied to each fruit’s mass, so it is included by multiplying that same observation by 1.08. If the whole bag is sold at $2.90 per kg, its price is , in dollars. Then and . Both unit conversion and percentage change square their multiplier in a variance.
Comparing Normal Distribution Sketches
Compare the centres, spreads and shaded areas
Battery lives, in months, have distributions and . Sketch both on one diagram. Then shade the probability that a B battery lasts between 21 and 29 months.
Put each peak above its mean. A is centred farther right. Its standard deviation is twice B’s, so its curve is wider and its peak is half as high. Each entire density has area one.
Use the standard deviations to mark the spread:
| Life | Mean | Mean ± SD |
|---|---|---|
| A | 30 | 22 to 38 |
| B | 25 | 21 to 29 |
On B’s curve, shade between the two cutoffs, rather than either tail:
For a normal distribution, about 68% lies within one standard deviation of the mean, 95% within two, and 99.7% within three. Use this to check the scale of a sketch; calculate the requested probability using the actual cutoffs.
Before you finish
State the parameters of every new distribution. Check the units, whether observations are independent, whether a coefficient multiplies one observation or counts several, and whether “difference” requests one tail or two. For sketches, label the actual mean and cutoffs; a larger standard deviation gives a wider, lower density curve.