Experimental probability, fairness and expected outcomes · Higher
Introduit| English | Sens |
|---|---|
| relative frequency/ˈrelətɪv ˈfriːkwənsi/ | fréquence relative |
Kerrich's 10 000 tosses
- During the Second World War, the mathematician John Kerrich was held in a camp in Denmark. With time to spare, he tossed a coin 10 000 times and recorded 5067 heads. His record became a classic example of how a relative frequency 相对频率 settles down.
- Higher questions use estimates to judge fairness and to predict.
Estimating a probability
- 18 reds in 60 spins: estimated $P(\text{red}) = 0.3$, so expect $0.3 \times 200 = 60$ reds in 200 spins.
- If the outcomes are red, blue, green with $0.3$, $0.45$ and $p$: $p = 0.25$.
Variation and sample size
- Short runs vary a lot: 6 heads in 10 tosses is 0.6. 502 in 1000 is 0.502.
- Combine results from several groups to get a better estimate: more trials, less variation.
An estimate from 1000 trials is generally more reliable than one from 10 trials.
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
Less variation in long runs.
Worked example: combining estimates
Three groups test a spinner. Group A: 18 reds in 60. Group B: 25 in 100. Group C: 47 in 140. Give the best estimate of P(red).
- Combine: total reds $18 + 25 + 47 = 90$; total spins $60 + 100 + 140 = 300$.
- Estimate: $\dfrac{90}{300} = 0.3$.
- Why: it uses all 300 spins, so it is more reliable than any one group (0.30, 0.25, 0.34).
Experimental probability, fairness and expected outcomes
Record each trial consistently and total the counts
Classify the worked-case statements, then explain the units or invariant that justifies each decision.
Group A gets 12 sixes in 50 rolls; Group B gets 38 in 250. Give the best estimate of P(six).
Difficulté 2/5 · 140 pièces + un petit bonus lors de la première réussite aujourd'hui
50 ÷ 300.
Using estimates in calculations
- With P(red) $= 0.3$, two independent spins: P(two reds) $= 0.09$; P(at least one red) $= 1 - 0.7^2 = 0.51$.
- Expected number of reds in 50 spins: $0.3 \times 50 = 15$.
P(a bus is late) is estimated at 0.15. How many late buses would you expect in 220 journeys?
Difficulté 2/5 · 140 pièces + un petit bonus lors de la première réussite aujourd'hui
0.15 × 220.
P(red) = 0.3. A spinner is spun 3 times. Find P(no reds) to 3 d.p.
Difficulté 2/5 · 140 pièces + un petit bonus lors de la première réussite aujourd'hui
$0.7^3$.
Theory against experiment
- A fair six-sided dice gives $\dfrac{1}{6}$ for each face. 600 rolls should give about 100 of each.
- Kerrich's 5067 heads is 0.5067 — close to 0.5, so the coin was very likely fair.
A coin gives 5067 heads in 10 000 tosses. What is the best conclusion?
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
0.5067 is very close to 0.5.
Do not average the groups' estimates when sample sizes differ. The mean of 0.30, 0.25 and 0.34 is 0.297. Pooling counts gives 0.3, weighting each spin equally. Add the counts, then divide.
The best combined estimate is always the mean of the groups' relative frequencies.
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
Pool the counts.
Estimate a probability by relative frequency, pooling all trials. More trials give a more reliable estimate. Use the estimate in expected values and probability calculations, and compare it with theory to judge fairness.
A biased spinner has P(red) = 0.3, P(blue) = 0.45, P(green) = 0.25. It is spun twice. Find P(both the same colour), and the expected number of greens in 80 spins.
Using estimates.