I.

Basic probability formulas

● P(A ⋃ B) = P(A) + P(B) - P(A ⋂ B)
𝑃(𝐴 ⋂ 𝐵)
● P(A | B) =
𝑃(𝐵)

𝑃(𝐵 | 𝐴) . 𝑃(𝐴)
● P(A | B) =
𝑃(𝐵)
● If A, B independent: P(A ⋂ B) = P(A) . P(B)

II. Discrete random variables

● ℳ = E(x) = ∑ xi . P(x=xi)

● σ2 = V(x) = ∑ (xi - ℳ)2 . P(x=xi)

= ∑ xi2 . P(x=xi) - ℳ2
● E(ax + by) = a.E(x) + b.E(y)
● V(ax + by) = a2 . V(x) + b2 . V(y)
● Probability mass function: f(xi) = P(x=xi)
● Cumulative distribution function: F(xi) = P(x≤xi)
● Some special distribution:
1. Discrete uniform distribution
1
○ P(x=Xi) =
𝑛

𝑎+𝑏
○ ℳ= 2
2
2 (𝑏 − 𝑎 + 1) − 1
○ σ =
12
2. Binomial distribution
○ P(x=k) = nCk . pk . (1-p)n-k

○ ℳ = n.p
○ σ2 = n.p . (1-p)
3. Poisson distribution
−λ.𝑇
𝑒
○ P(x=k) = (λ.T)k
𝑘!

○ ℳ = λ.T
○ σ2 = λ.T
 4. Hypergeometric distribution
𝐾𝐶𝑘 . (𝑁−𝐾)𝐶(𝑛−𝑘)
○ P(x=k) =
𝑁𝐶𝑛

○ ℳ = n.p
𝑁−𝑛
○ σ2 = n.p.(1-p). 𝑁 − 1

5. Geometric distribution
○ P(x=k) = (1-p)k-1 . p
1
○ ℳ= 𝑝

1−𝑝
○ σ2 = 2
𝑝
6. Negative binomial distribution
○ P(x=k) = (k-1)C(r-1) . pr . (1-p)k-r
𝑟
○ ℳ= 𝑝

𝑟 . (1 − 𝑝)
○ σ2 = 2
𝑝

III. Continuous random variable

𝑏
● Probability density function f(x): P(a<x<b) = ∫f(x) dx
𝑎

● Cumulative distribution function F(x):

○ F(xi) = P(x≤xi)
○ F(xi)’ = f(x i)
+∞
● ℳ = E(x) = ∫ 𝑥. 𝑓(𝑥) 𝑑𝑥
−∞

+∞
𝑛
● E(xn) = ∫ 𝑥 . 𝑓(𝑥) 𝑑𝑥
−∞

+∞
2 2 2
● σ = V(x) = ∫ 𝑥 . 𝑓(𝑥) 𝑑𝑥 - ℳ
−∞

● Some special distribution:

1. Continuous uniform distribution
1
○ f(x) = , a≤x≤b
𝑏−𝑎
 = 0 , elsewhere
𝑎+𝑏
○ ℳ= 2
2
2 (𝑏 − 𝑎)
○ σ =
12

2. Normal distribution N(ℳ, σ2)

𝑥−ℳ
○ z=
σ
2
𝑥
1 2
○ f(z) = . 𝑒
2Π
○ ϕ(x) = p(z<xi)
○ ϕ(-x) = 1 - ϕ(x)
3. Normal distribution approximate binomial and poisson distribution
a. Binomial (np > 5 and n(1-p) > 5)
𝑥 − 𝑛.𝑝
■ z=
𝑛.𝑝.(1−𝑝)
■ P(XBINORM ≤ a) = P(XNORMAL ≤ a+0.5)
■ P(XBINORM ≧ a) = P(XNORMAL ≧ a-0.5)
b. Poisson
𝑥−λ
■ z=
λ
■ P(XPOISSON ≤ a) = P(XNORMAL ≤ a+0.5)
■ P(XPOISSON ≧ a) = P(XNORMAL ≧ a-0.5)

4. Exponential distribution
−λ.𝑇
○ f(x) = λ . 𝑒 , x>0

○ = 0 , elsewhere
−λ.𝑎
○ P(x ≧ a) = 𝑒 ,(a > 0)
1
○ ℳ= λ

1
○ σ2 = 2
λ

IV. Descriptive statistic (Take a sample of size n from population N)

 ∑ 𝑥𝑖
● Sample mean: 𝑥 =
𝑛

𝑛+1 𝑥𝑐𝑒𝑖𝑙(𝐿) + 𝑥𝑓𝑙𝑜𝑜𝑟(𝐿)

● Sample median: L = so Median =
2 2
● Mode: Số phần tử xuất hiện nhiều nhất
● Range: max - min
2
∑ ( 𝑥 − 𝑥𝑖)
● Sample variance: s2 =
𝑛−1
● Quatiles:
𝑥𝑐𝑒𝑖𝑙(𝐿 ) + 𝑥𝑓𝑙𝑜𝑜𝑟(𝐿 )
𝑛+1 1 1
○ L1 = so Q1 =
4 2

𝑥𝑐𝑒𝑖𝑙(𝐿 ) + 𝑥𝑓𝑙𝑜𝑜𝑟(𝐿 )
𝑛+1 2 2
○ L2 = so Q2 =
2 2

𝑥𝑐𝑒𝑖𝑙(𝐿 ) + 𝑥𝑓𝑙𝑜𝑜𝑟(𝐿 )
3.(𝑛 + 1) 3 3
○ L3 = so Q3 =
4 2

V. Sampling distribution

● Population mean ℳ, variance σ2. Sample size n. (Normal distribution or n > 30):
2
σ
○ Phân phối của 𝑋 có dạng: N(ℳ , )
𝑛
2 2
σ1 σ2
○ Phân phối của 𝑋 - 𝑋 có dạng: N(ℳ1 - ℳ2 , 𝑛 + 𝑛 )
1 2
1 2

● For proportion of population p, sample size n. (np ≧ 5 or n.(1-p) ≧ 5):

𝑃.(1 − 𝑃)
○ Phân phối của 𝑃 có dạng: N(𝑃 , )
𝑛

𝑃1.(1 − 𝑃1) 𝑃2.(1 − 𝑃2)

○ Phân phối của 𝑃 - 𝑃 có dạng: N(𝑃 - 𝑃 , + )
1 2 1 2 𝑛1 𝑛2

VI. Statistical intervals - Test claims for one sample

● (l, u) = (𝑋 - E, 𝑋 + E)
● width = 2E
● P-value = 2 . P(Z > |Z0|)
 1. Population variance known
σ
○ E=𝑧
α/2
.
𝑛

𝑋−ℳ
○ 𝑧0 =
σ/ 𝑛
2. Population variance unknown
○ n > 30:
𝑆
■ E=𝑧
α/2
.
𝑛

𝑋−ℳ
■ 𝑧0 =
𝑆/ 𝑛
○ n ≤ 30:
𝑆
■ E=𝑡
α/2, 𝑛−1
.
𝑛

𝑋−ℳ
■ 𝑡0 =
𝑆/ 𝑛
● For propotion:

○ (l, u) = (𝑃 - E, 𝑃 + E)

𝑃.(1 − 𝑃)
○ E=𝑧 .
α/2 𝑛

𝑃−𝑃
○ 𝑧0 =
𝑃.(1 − 𝑃)
𝑛

○ Nếu đề không cho 𝑃, mặc định 𝑃 = 0.5

● Nếu là one-side thì tương tự nhưng thay α/2 thành α

VII. Test claims for 2 samples (2 population independent, normal distribution or both n1, n2 > 30)

● (l, u) = (𝑋 - 𝑋 - E , 𝑋 - 𝑋 + E)
1 2 1 2

1. Population variance known

2 2
σ1 σ2
○ E=𝑧
α/2
.
𝑛1
+ 𝑛2
 𝑋1 − 𝑋2 − ∆0
○ 𝑧0 =
2 2
σ1 σ2
𝑛1
+ 𝑛2

2. Population variance unknown

2
○ Assume σ
1
= σ22

■ Degree of freedom: df = 𝑛 + 𝑛 + 2
1 1
2 2
(𝑛1 − 1) . 𝑆1 + (𝑛2 − 1) . 𝑆2
2
■ 𝑆𝑝 =
𝑛1+ 𝑛2 − 2

2 2
𝑆𝑝 𝑆𝑝
■ E=𝑡
α/2, 𝑑𝑓
.
𝑛1
+ 𝑛2

𝑋1 − 𝑋2 − ∆0
■ 𝑡0 =
2 2
𝑆𝑝 𝑆𝑝
𝑛1
+ 𝑛2

2 2
○ Not assume σ = σ
1 2
2

( )
2 2
𝑆1 𝑆2
𝑛1
+ 𝑛2
■ Degree of freedom: df = 4 4
𝑆1 𝑆2
2 + 2
𝑛1 . (𝑛1 − 1) 𝑛2 . (𝑛2 − 1)

2 2
𝑆1 𝑆2
■ E=𝑡
α/2, 𝑑𝑓
.
𝑛1
+ 𝑛2

𝑋1 − 𝑋2 − ∆0
■ 𝑡0 =
2 2
𝑆1 𝑆2
𝑛1
+ 𝑛2

● For propotion:

○ (l, u) = (𝑃 - 𝑃 - E , 𝑃 - 𝑃 + E)
1 2 1 2

𝑃1 . (1 − 𝑃1) 𝑃2 . (1 − 𝑃2)
○ E=𝑧
α/2
.
𝑛1
+ 𝑛2
 𝑥1 + 𝑥2
○ 𝑃= (trong đó xi = n . 𝑃 )
𝑛1 + 𝑛2 𝑖

𝑃1 − 𝑃2 − ∆0
○ 𝑧0 =
𝑃 . (1 − 𝑃) . ( 1
𝑛1
+
1
𝑛2 )
VIII. Linear Regression

● SXY = ∑(𝑥𝑖 − 𝑥)(𝑦𝑖 − 𝑦) = ∑𝑥𝑖𝑦𝑖 − 𝑛. 𝑥. 𝑦

2
( ) = ∑𝑥 2 2
● SXX = ∑ 𝑥 − 𝑥 − 𝑛.𝑥
𝑖 𝑖

2
= ∑(𝑦 − 𝑦) = ∑𝑦
2 2
● SYY − 𝑛.𝑦
𝑖 𝑖

𝑆𝑋𝑌 ∑𝑥𝑖𝑦𝑖 − 𝑛 . 𝑥 . 𝑦
● Slope: β = = 2
1 𝑆𝑋𝑋 2
∑𝑥𝑖 − 𝑛 . 𝑥

● Intercept: β = 𝑦 - β . 𝑥
0 1
2
● Error sum of square: SSE = ∑ 𝑦𝑖 − 𝑦𝑖 ( )
2
● Regression sum of square: SSR = ∑ 𝑦𝑖 − 𝑦 ( )
2
● Total sum of square: SST = ∑ 𝑦𝑖 − 𝑦 ( )
● SSE + SSR = SST

𝑆𝑆𝐸
● Standard error: σ =
𝑛−2

𝑆𝑆𝑅 𝑆𝑋𝑌
● Coefficient of correlation: R = =
𝑆𝑆𝑇 𝑆𝑋𝑋 . 𝑆𝑌𝑌

● Test claims about the slope (df = n-2):

2
σ
○ se(β ) =
1 𝑆𝑋𝑋

β1 − β1,0
○ 𝑡0 =
𝑠𝑒(β1)
● Test claims about the intercept (df = n-2):

( )
2 2
1 𝑥
○ se(β ) =
0
σ . 𝑛
+ 𝑆𝑋𝑋

β0 − β0,0
○ 𝑡0 =
𝑠𝑒(β0)

𝑅−0
● Test claims about the coefficient of correlation (df = n-2): 𝑡 =
0 2
1−𝑅
𝑛−2