Binomial, normal and Poisson models
Introducido| English | Significado |
|---|---|
| normal distribution/ˈnɔːml ˌdɪstrɪˈbjuːʃn/ | A continuous, symmetric bell-shaped probability distribution specified by its mean and standard deviation. |
| discrete random variable/dɪˈskriːt ˈrændəm ˈveərɪəbl/ | A numerical outcome of a random experiment with a finite or countably infinite set of possible values. |
| cumulative distribution function/ˈkjuːmjʊlətɪv ˌdɪstrɪˈbjuːʃn ˈfʌŋkʃn/ | A function giving the probability that a random variable is less than or equal to a specified value. |
5,738 soldiers' chests, 1846
- In 1846, the Belgian statistician Adolphe Quetelet published the chest sizes of 5,738 Scottish soldiers. Plotted, the counts formed a smooth bell-shaped curve around the average — one of the first real data sets shown to follow the normal distribution 正态分布.
- S1 builds models for random quantities: a discrete random variable 离散型随机变量 that takes separate values, and the normal distribution for continuous measurements.
A probability distribution
- A discrete random variable $X$ has a table of values and probabilities that add to 1. For example, $\mathrm{P}(X = 0) = 0.2$, $\mathrm{P}(X = 1) = 0.5$, $\mathrm{P}(X = 2) = 0.3$.
- The cumulative distribution function 累积分布函数 is $F(x) = \mathrm{P}(X \le x)$: here $F(1) = 0.7$.
$\mathrm{P}(X = x) = kx$ for $x = 1, 2, 3, 4$. Find $k$.
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
$k(1 + 2 + 3 + 4) = 1$, so $k = 0.1$.
Mean and variance
- $\mathrm{E}(X) = \sum x\,\mathrm{P}(X = x)$ and $\mathrm{Var}(X) = \mathrm{E}(X^2) - [\mathrm{E}(X)]^2$.
- For the table above: $\mathrm{E}(X) = 0.5 + 0.6 = 1.1$, $\mathrm{E}(X^2) = 0.5 + 1.2 = 1.7$, so $\mathrm{Var}(X) = 1.7 - 1.21 = 0.49$.
Binomial, normal and Poisson models
For a discrete random variable, sum the probabilities to 1 and calculate E(X)=sum xP(X=x)
Compare the model with the worked case and explain one change.
A favor $\mathrm{P}(X = x) = 0.1x$, $x = 1, 2, 3, 4$, find $\mathrm{E}(X)$.
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
$0.1(1 + 4 + 9 + 16) = 3$.
For the same $X$, $\mathrm{E}(X^2) = 10$. Find $\mathrm{Var}(X)$.
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
$10 - 3^2 = 1$.
Worked example: a coded variable
Using the distribution above, a game pays $Y = 3X - 1$ yuan. Find $\mathrm{E}(Y)$, $\mathrm{Var}(Y)$ and $\mathrm{P}(Y > 0)$.
- Mean: $\mathrm{E}(aX + b) = a\mathrm{E}(X) + b$, so $\mathrm{E}(Y) = 3(1.1) - 1 = 2.3$.
- Variance: $\mathrm{Var}(aX + b) = a^2\mathrm{Var}(X)$, so $\mathrm{Var}(Y) = 9 \times 0.49 = 4.41$ — adding 1 changes nothing.
- Probability: $Y > 0$ means $3X > 1$, so $X \ge 1$: $\mathrm{P}(Y > 0) = 0.5 + 0.3 = 0.8$.
$\mathrm{E}(X) = 4$ y $\mathrm{Var}(X) = 2$. Find $\mathrm{E}(5 - 2X)$ y $\mathrm{Var}(5 - 2X)$.
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
$5 - 8 = -3$; $(-2)^2 \times 2 = 8$. Variance is never negative.
The normal distribution
- $X \sim \mathrm{N}(\mu, \sigma^2)$ is symmetric about $\mu$; about 68% of values lie within $\sigma$ of $\mu$ and about 95% within $2\sigma$.
- Standardise with $Z = \dfrac{X - \mu}{\sigma}$: if heights are $\mathrm{N}(170, 8^2)$ cm, then $\mathrm{P}(X > 182) = \mathrm{P}(Z > 1.5) = 0.0668$.
$X \sim \mathrm{N}(500, 6^2)$ grams. Find $\mathrm{P}(X > 510)$, to 4 d.p.
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
$z = \tfrac{10}{6} = 1.667$, and $\mathrm{P}(Z > 1.667) = 0.0478$.
Working backwards
- For "the tallest 10%", find $z$ with $\mathrm{P}(Z > z) = 0.1$: $z = 1.2816$, so $x = 170 + 1.2816 \times 8 = 180.3$ cm.
- To find an unknown $\sigma$: if $\mu = 170$ and $\mathrm{P}(X < 160) = 0.1$, then $\dfrac{160 - 170}{\sigma} = -1.2816$, so $\sigma = 7.80$ cm.
$X \sim \mathrm{N}(60, \sigma^2)$ y $\mathrm{P}(X > 70) = 0.0228$. Find $\sigma$.
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
$\mathrm{P}(Z > 2) = 0.0228$, so $\tfrac{10}{\sigma} = 2$.
Writing $\mathrm{Var}(3X - 1) = 3\mathrm{Var}(X) - 1$. Variance measures spread in squared units, so a scale factor is squared and a shift has no effect: $\mathrm{Var}(3X - 1) = 9\mathrm{Var}(X)$. Only the mean follows $a\mathrm{E}(X) + b$.
A discrete distribution's probabilities add to 1; $\mathrm{E}(X) = \sum xp$ and $\mathrm{Var}(X) = \mathrm{E}(X^2) - [\mathrm{E}(X)]^2$, with $\mathrm{E}(aX + b) = a\mathrm{E}(X) + b$ and $\mathrm{Var}(aX + b) = a^2\mathrm{Var}(X)$. For $\mathrm{N}(\mu, \sigma^2)$, standardise with $Z = \tfrac{X - \mu}{\sigma}$, and use the inverse table to work back to $x$, $\mu$ or $\sigma$.
The masses of apples are normally distributed. 15% weigh less than 140 g and 10% weigh more than 200 g. Find the mean and standard deviation.
Two z-values, two equations.