Memo (I) Probability and Statistics

Let (Ω, A, P) be a probability space.



 0 , x < x1 3 Some usual distribution
 i
 X
2. c.d.f. : FX (x) = P(X = xi ) , xi ≤ x < xi+1 , Bernoulli distribution B(p), p ∈ (0, 1)
1 Probability 

 j=1
, x ≥ xn P(X = 1) = p, P(X = 0) = 1 − p = q,

1
For A, B ∈ A:
E(X) = p, V(X) = pq and ϕX (t) = q + peit , t ∈ R.
n
1. P(Ā) = 1 − P(A), X
3. Expectation : E(g(X)) = g(xi )P(X = xi ), g piecewise cont. Binomial distribution B(n, p), n ∈ N∗ , p ∈ (0, 1)
2. A ⊆ B =⇒ P(A) ≤ P(B), i=1

3. P(A \ B) = P(A) − P(A ∩ B), n

X P(X = k) = Cnk pk q n−k with 1 − p = q, and Cnk = k!(n−k)!
n!
.
4. P(A ∪ B) = P(A) + P(B) − P(A ∩ B). 4. Charac. function : ϕX (t) = eitxj P(X = xj ), t ∈ R it n
E(X) = np, V(X) = npq and ϕX (t) = (q + pe ) , t ∈ R.
j=1

Conditional probability Poisson distribution P(λ), λ > 0

Continuous case: X(Ω) = I, I interval of R
1. For A, B ∈ A with P(B) 6= 0. The conditional probability of A λk −λ
fX (x)dx, B ∈ B(R) .
R
1. Distribution : P(X ∈ B) = B
P(X = k) = e
P(B ∩ A) k! it
given B is P(A|B) = . Z E(X) = λ, V(X) = λ, and ϕX (t) = eλ(e −1) , t ∈ R.
P(B)
2 p.d.f.: fX : R → R+ satisfying fX (x)dx = 1.
2. [Law of total probability] Let B1 , B2 , . . . , Bn , be events in A, R Uniform distribution U(a, b), a < b real numbers
with P(Bi ) 6= 0, such that ∪n
i=1 Bi = Ω and Bi ∩ Bj = ∅ for i 6= j . Rx d
n 3. c.d.f. : FX (x) = fX (t)dt and fX = F , a.e..
dx X
−∞ 1 a+b (b−a)2
X f (x) = 1 (x), x ∈ R, E(X) = , V(X) =
Then, for any event A in A: P(A) = P(A|Bi )P(Bi ). Z b−a [a,b] 2 12

i=1
4. Expectation : E(g(X)) = g(x)fX (x)dx, g mes. function Exponential distribution E(λ)
R
Independence
eitx fX (x)dx, t ∈ R f (x) = λe−λx 1[0,+∞[ (x), x ∈ R,
R 1 1
Two events A, B ∈ A with P(B) 6= 0 are independent if 5. Charac. function : ϕX (t) = R E(X) = λ
, V(X) = λ2

P(A|B) = P(A) or equivalently P(A ∩ B) = P(A)P(B) .

Same distribution Two random variables X, Y have the same Normal distribution N(m, σ )nm ∈ R, σ > 0
2

probability distribution if FX = FY or ϕX = ϕY . 1
2 Real random variable (R) f (x) =
1 x−m 2
√ e− 2 ( σ ) , x ∈ R,
σ 2π
1. The probability distribution of a real random variable X is Properties of expectation and variance −t2 σ 2
E(X) = m, V(X) = σ 2 and ϕX (t) = eitm e 2 ,t ∈ R
1. The expectation is linear i.e. for λ, µ ∈ R, we have
given by P(X ∈ B) = P({ω ∈ Ω; X(ω) ∈ B}), B ∈ B(R).
E(λX + µY ) = λE(X) + µE(Y ).
X −m
2. The c.d.f. of a random variable X is Moreover X ∼ N(m, σ 2 ) ⇔ Z = ∼ N(0, 1).
σ
2. The variance of X is defined by V(X) = E((X − E(X))2 ). We
FX : R → [0, 1] defined by FX (x) = P(X ≤ x), x ∈ R. Given α ∈ [0, 1] then there is zα > 0 s.t.
have V(X) = E(X 2 ) − (E(X))2 and the variance is quadratic i.e. P(|Z| ≤ zα ) = 1 − α ⇔ P(Z ≤ zα ) = 1 − α2 .
It satisfies P(a < X ≤ b) = FX (b) − FX (a), a, b ∈ R, a < b .
V(λX + µ) = λ2 V(X) for λ, µ ∈ R.
Chi-square distribution χ2 (n), n ∈ N∗
p
Discrete case: X(Ω) = {x1 , . . . , xn } ⊂ R or X(Ω) = 3. The standard deviation of X is σ(X) = V(X). n
x 2 −1 e− 2
x

{x1 , . . . , xn , . . . } ⊂ R f (x) = n 1]0,+∞] (x), x ∈ R, E(X) = n, V(X) = 2n

4. [Chebyshev’s inequality] Given a random variable X with finite 2 2 Γ( n2 )
1. Distribution : P(X = xi ) = P({ωi ; X(ωi ) = xi }), i = 1, . . . , n . and ϕX (t) = (1 − 2it)−n/2 , ∀t ∈ R.
V(X)
mean and variance, ∀ε > 0, P(|X − E(X)| ≥ ε) ≤ ε2
.

Centrale Nantes / M1-ENG-CM Page 1/2 2022-2023

4 Real random vector (Rn ) We have Cov(X, X) = V(X) . 3. If X1 , . . . , Xn are i.i.d. with Xi ∼ N(mi , σi2 ), then n
P
i=1 λi Xi ∼
2. If the random variables X1 , . . . , Xn are jointly distributed then N(m, σ ), where P
2
λi is an arbitrary real number and m =
1. The probability distribution of a random vector X is
P n 2 n 2 2
V(X1 + · · · + Xn ) = n i=1 λi mi , σ = i=1 λi σi .
P P
i=1 V(Xi ) + i6=j Cov(Xi , Xj ).
PX (B) = P(X ∈ B) = P({ω ∈ Ω; X(ω) ∈ B}), B ∈ B(Rn ). 4. If X is N(0, 1) , then X 2 ∼ χ2 (1) .
It is called the joint distribution of X1 , . . . , Xn . Let A = A1 × · · · × An ⊂ R with Ai ∈ B(R), i = 1, . . . , n.
n
5. If X1 , . . . , Xk are i.i.d. with Xi is χ2 (ni ), then X1 + · · · + Xk ∼
χ2 (n1 + · · · + nk ).
2. The c.d.f. of a random vector X is FX : Rn → [0, 1] defined for Conditional distributions
all x = (x1 , . . . , xn ) ∈ Rn by FX (x) = P(X1 ≤ x1 , . . . , Xn ≤ xn ), 1. Given two discrete random vectors X and Y such that
P(Y = y) 6= 0. The distribution of X given Y is
6 Convergence
3. The marginal distributions of a random vector X are the dis- P(X = x, Y = y) Let {Xn }n∈N be a sequence of random variables.
tributions PXi of the random variables Xi , i = 1, . . . , n. P(X = x|Y = y) = .
P(Y = y)
P 1. Convergence in probability
We have P(X ∈ A|Y = y) = x∈A P(X = x|Y = y).
Discrete case: X(Ω) = {(xi1 , . . . , xin ), ij ∈ N, xij ∈ R} P
2. Given two continuous random vectors X and Y such Xn −→ X ⇔ ∀ε > 0, lim P(|Xn − X| > ε) = 0.
n→+∞
1. Distribution: P(X1 = xi1 , . . . , Xn = xin ) .
that fY (y) = 6 0. The conditional p.d.f. of X given Y is
2. Convergence almost surely
2. Expectation: fX,Y (x, y)
fX|Y =y (x) = . a.s.
Xn −→ X ⇔ P({ω ∈ Ω : lim Xn (ω) 6= X(ω)}) = 0.
X fY (y) n→+∞
E(g(X)) = g(xi1 , . . . , xin )P(X1 = xi1 , . . . , Xn = xin ) R
(xi1 ,...,xin )∈X(Ω) We have P(X ∈ A|Y = y) = A fX|Y =y (x)dx1 . . . dxn . 3. Convergence in distribution
L
3. Marginal dist.: Xn −→ X ⇔ lim FXn (t) = FX (t), for t where FX is cont, or
Independence Let X1 , . . . , Xn be random variables. n→+∞
X
P(Xk = xik ) = P(X1 = xi1 , . . . , Xn = xin )
if and only if: lim ϕXn (t) = ϕ(t), ∀t ∈ R.
xij 6=xi n→+∞
k 1. X1 , . . . , Xn are independent if, for all A1 , . . . , An ,
P(X1 ∈ A1 , . . . , Xn ∈ An ) = n
Q
i=1 P(Xi ∈ Ai ). 4. Convergence in moment of second order
Continuous case: X(Ω) box of Rn L2
Xn −→ X ⇔ lim E(|Xn − X|2 ) = 0.
Z 2. If X1 , . . . , Xn are independent, g1 (X1 ), . . . , gn (Xn ) are also n→+∞
1. Distribution: P(X ∈ B) = fX (x1 , . . . , xn )dx1 . . . dxn . independent for all measurable function gi : R → R, i = 1, . . . , n.
B
Z 3. If X1 , . . . , Xn are discrete, they are independent Important results
n
2. p.d.f.: fX : R → R+ s.t. fX (x1 , .., xn )dx1 ..dxn = 1 1. Given a continuous function f : R → R then
⇔ P(X1 = xi1 , . . . , Xn = xin ) = P(X1 = xi1 ) . . . P(Xn = xin ).
Rn a.s. a.s.
R x1 R xn Xn −→ X ⇒ f (Xn ) −→ f (X).
3. c.d.f.: FX (x1 , . . . , xn ) = −∞
··· −∞
fX (t1 , . . . , tn )dt1 . . . dtn 4. If X1 , . . . , Xn are continuous, they are independent
⇔ fX (x1 , . . . , xn ) = fX1 (x1 ) . . . fXn (xn ). 2. [Slutsky’s theorem] Given two sequences of random variables
∂n
and fX = F
∂x1 ...∂xn X
a.e. a.s. L L
{Xn }n∈N and {Yn }n∈N then Xn → c, Yn → Y ⇒ Xn Yn → cY.
5. If X1 , . . . , Xn are independent then:
4. Expectation:
Z 1. E(X1 × · · · × Xn ) = E(X1 ) × · · · × E(Xn ), 3. [Central Limit Theorem of Laplace] Let {Xi }i=1,n , i.i.d.
E(g(X)) = g(x1 , . . . , xn )fX (x1 , . . . , xn )dx1 . . . dxn with mean m and variance σ 2 . Then, if X n = X1 +···+X n
,
Rn
2. Cov(Xi , Xj ) = ρ(Xi , Xj ) = 0 pour i 6= j, n

3. V(X1 + · · · + Xn ) = V(X1 ) + · · · + V(Xn ) Xn − m L

5. Marginal dist. : defined through its marginal p.d.f Zn = √ −→ Z ∼ N(0, 1).
Z σ/ n
fXk (xk ) = fX (x1 , . . . , xn )dx1 . . . dxk−1 dxk+1 . . . dxn
Rn−1
5 Function of random variables 4. [Weak law of large number] Let {Xi }i=1,n , i.i.d. with mean
X1 +···+Xn P
1. Let X1 , . . . , Xk i.i.d. with Xi ∼ B(ni , p), then X1 + · · · + Xk ∼ and variance then X n = n
→ E(X1 ).
Covariance B(n1 + · · · + nk , p).
1. The covariance of two random variables X, Y is 5. [Strong law of large number] Let {Xi }i=1,n , i.i.d. with mean
2. Let X1 , . . . , Xn i.i.d. with Xi ∼ P(λi ), then X1 + · · · + Xn ∼
a.s.
Cov(X, Y ) = E(X − E(X))(Y − E(Y )) = E(XY ) − E(X)E(Y ). P(λ1 + · · · + λn ). and variance, moreover if E(|Xi |) < ∞ then X n → E(X1 ).

Centrale Nantes / M1-ENG-CM Page 2/2 2022-2023