The document summarizes several discrete probability distributions including:
1. Uniform distribution which assigns equal probability to integers between two values.
2. Binomial distribution which models the number of successes in independent yes/no experiments.
3. Poisson distribution which models the number of rare, independent events occurring in an interval of time or space.
4. Multinomial distribution which generalizes the binomial to experiments with more than two possible outcomes.
The document summarizes several discrete probability distributions including:
1. Uniform distribution which assigns equal probability to integers between two values.
2. Binomial distribution which models the number of successes in independent yes/no experiments.
3. Poisson distribution which models the number of rare, independent events occurring in an interval of time or space.
4. Multinomial distribution which generalizes the binomial to experiments with more than two possible outcomes.
The document summarizes several discrete probability distributions including:
1. Uniform distribution which assigns equal probability to integers between two values.
2. Binomial distribution which models the number of successes in independent yes/no experiments.
3. Poisson distribution which models the number of rare, independent events occurring in an interval of time or space.
4. Multinomial distribution which generalizes the binomial to experiments with more than two possible outcomes.
The document summarizes several discrete probability distributions including:
1. Uniform distribution which assigns equal probability to integers between two values.
2. Binomial distribution which models the number of successes in independent yes/no experiments.
3. Poisson distribution which models the number of rare, independent events occurring in an interval of time or space.
4. Multinomial distribution which generalizes the binomial to experiments with more than two possible outcomes.
Distribution Parameters Possible Description Range ΩX E[X] Var(X) k ∈ ΩX ) X ∼ Unif(a, b) Equally likely to be any a+b (b − a)(b − a + 2) 1 Uniform (disc) for a, b ∈ Z {a, . . . , b} integer in [a, b] 2 12 b−a+1 and a ≤ b 1−k X ∼ Ber(p) Takes value 1 with prob p pk (1 − p) Bernoulli {0, 1} p p(1 − p) for p ∈ [0, 1] and 0 with prob 1 − p X ∼ Bin(n, p) Sum of n iid Ber(p) rvs. # n n−k pk (1 − p) Binomial for n ∈ N, of heads in n independent {0, 1, . . . , n} np np(1 − p) k
and p ∈ [0, 1] coin flips with P (head) = p.
# of events that occur in λk Poisson X ∼ Poi(λ) one unit of time {0, 1, . . .} λ λ e−λ for λ > 0 independently with rate λ k! per unit time # of independent Bernoulli k−1 X ∼ Geo(p) 1 1−p (1 − p) p Geometric trials with parameter p up to {1, 2, . . .} for p ∈ [0, 1] p p2 and including first success # of successes in n draws K N −K
X ∼ HypGeo(N, K, n) (w/o replacement) from N {max (0, n + K − N ), K K (N − K) (N − n) k n−k Hypergeometric for n, K ≤ N n n N
items that contain K . . . , min (n, K)} N N 2 (N − 1) n and n, K, N ∈ N successes in total Sum of r iid Geo(p) rvs. # k−1 k−r Negative X ∼ NegBin(r, p) r r (1 − p) pr (1 − p) of independent flips until rth {r, r + 1, . . .} r−1 Binomial for r ∈ N, p ∈ [0, 1] p p2 head with P (head) = p X ∼ Multr (n, p) Generalization of the ki ∈ {0, . . . , n}, np1 V ar(Xi ) = npi (1 − pi ) for r, n ∈ N and Binomial distribution, n n Qr Multinomial i ∈ {1, . . . , r} . Cov(Xi , Xj ) = k1 ,...,kr i=1 pki i p =P(p1 , p2 , ..., pr ), trials with r categories each E[X] = np = .. r and Σki = n −npi pj , i 6= j i=1 pi = 1 with probability pi npr Generalization of the X ∼ MVHGr (N, K, n) V ar(Xi ) = Hypergeometric distribution, ki ∈ {0, . . . , Ki }, n KN1 N −Ki N −n nK Qr Ki N · · N −1
Multivariate for r, n ∈ N, i N i=1 ki n draws from r categories i ∈ {1, . . . , r} K . Hypergeometric K ∈ NrPand E[X] = n = .. Cov(Xi , Xj ) = N
r each with Ki successes and Σki = n N i Kj N −n n N = i=1 Ki −n K N N · N −1 , i 6= j (w/out replacement) n KNr University of Washington CSE312 Alex Tsun & Mitchell Estberg Continuous Distributions Possible CDF Distribution Parameters Range ΩX E[X] Var(X) PDF fX (x) for x ∈ ΩX Description (FX (x) = P(X ≤ x)) Equally likely to be 2 1 0 if x < a X ∼ Unif(a, b) a+b (b − a) x − a Uniform any real number in [a, b] b−a for a < b 2 12 if a ≤ x < b [a, b] b−a 1 if x ≥ b
Time until first (
X ∼ Exp(λ) 1 1 0 if x < 0 Exponential event in Poisson [0, ∞) λe−λx for λ > 0 λ λ2 1−e −λx if x ≥ 0 process X ∼ N (µ, σ 2 ) 2 ! 2 1 (x − µ) x−µ Normal for µ ∈ R, Standard bell curve (−∞, ∞) µ σ √ exp − Φ and σ 2 > 0 σ 2π 2σ 2 σ
Sum of r iid Exp(λ)
rvs. Time to rth X ∼ Gam(r, λ) event in Poisson r r λr r−1 −λx Note: Γ(r) = (r − 1)! Gamma [0, ∞) x e for r, λ > 0 process. Conjugate λ λ2 Γ(r) for integers r. prior for Exp, Poi parameter λ Conjugate prior for X ∼ Beta(α, β) α αβ Γ(α + β) α−1 β−1 Beta Ber, Bin, Geo, (0, 1) x (1 − x) for α, β > 0 α+β 2 (α + β) (α + β + 1) Γ(α)Γ(β) NegBin parameter p Generalization of X∼ Beta distribution. 1 Qr xai −1 , Dir(α1 , α2 , . . . , αr ) xi ∈ (0, 1); αi Dirichlet Conjugate prior for P r E[Xi ] = Pr B(α) Pir i=1 for αi , r > 0 and i=1 xi = 1 αj xi ∈ (0, 1), i=1 xi = 1 Multinomial j=1 r ∈ N, αi ∈ R parameter p X ∼ Nn (µ, Σ) 1 · Multivariate Generalization of (2π)n/2 |Σ|1/2 for µ ∈ Rn and Rn µ Σ 1 T −1 Normal Normal distribution exp(− 2 (x − µ) Σ (x − µ)) Σ ∈ Rn×n University of Washington CSE312 Alex Tsun & Mitchell Estberg