Several functional forms of f (xi ) are as follows.

Binomial distribution (二項分布):

⎧  
⎪
⎪ n
Discrete uniform distribution (離散型一様分布): ⎪
⎪
⎪ p x (1 − p)n−x , x = 0, 1, 2, · · · , n
⎧ ⎨
⎪
⎪ 1 f (x) = ⎪
⎪ x
⎪
⎨ N,
⎪ x = 1, 2, · · · , N ⎪
⎪
⎪
f (x) = ⎪ ⎩ 0, otherwise,
⎪
⎪
⎪
⎩ 0, otherwise
where 0 ≤ p ≤ 1 and n = 1, 2, · · ·.
where N = 1, 2, · · ·.

n
(a+b)n = nC x a
x n−x
b −→ Binomial Theorem (二項定理)
Bernoulli distribution (ベルヌイ分布):
⎧ x=0
⎪
⎪
⎪ x n
⎨ p (1 − p) , x = 0, 1
1−x
f (x) = ⎪ n!
⎪
⎪
⎩ 0, nC x = = n! = 1·2 · · · n (factorial of n)
otherwise x x!(n − x)!
where 0 ≤ p ≤ 1. X ∼ B(n, p)

32 33

Poisson distribution (ポアソン分布): ＜ Review ＞ e

⎧ −λ x Note that the definition of e is given by:
⎪
⎪
⎪ e λ
⎪
⎨ , x = 0, 1, · · ·
f (x) = ⎪ x!
⎪
⎪
⎪ 1
h
⎩ 0,
1
otherwise, e = lim(1 + x) x = lim 1 +
x→0 h→∞ h
where λ > 0. = 2.71828182845905.

Notation
exp(x) = e x

34 35

＜ Review ＞ Taylor series expansion about x0 Geometric distribution (幾何分布):


f (x0 ) f (x0 )
(k) ⎧
f (x) = f (x0 )+ f  (x0 )(x−x0 )+ (x−x0 )k +· ⎪
⎪
⎪ x
(x−x0 )2 +· · ·+
k! ⎨ p(1 − p) , x = 0, 1, · · ·
2! f (x) = ⎪
⎪
⎪
⎩ 0, otherwise,
where the kth derivative of f (x) is f (k) (x).
Taylor series expansion of f (x) = e x about x = 0 where 0 < p ≤ 1.

1 1 ∞
xk
e x = 1 + x + x2 + x3 + · · · = Negative binomial distribution (負の二項分布):
2! 3! k! ⎧ 
k=0
⎪
⎪ r + x − 1
⎪
⎪
⎪ pr (1 − p) x , x = 0, 1, · · ·
where f (k) (x) = e x . Set x = λ and k = x. ⎨
f (x) = ⎪
⎪ x
∞ ⎪
⎪
⎪
e−λ λ x ⎩ 0, otherwise,
1=
k=0
x!
where 0 < p ≤ 1 and r > 0.

36 37
Taylor series expansion of f (x) = (1 − x)−r about x = 0
r + k − 1
r(r + 1) 2 r(r + 1)(r + 2) 3
f (x) = 1+rx+ x+ x +· · ·+ xk +· ·
2! 3! k
r + k − 1 Hypergeometric distribution (超幾何分布):
where f (x) =
(k)
xk . ⎧ ⎛⎜ K ⎞⎟⎛⎜ M − K ⎞⎟
k ⎪
⎪
⎪ ⎜⎜⎜ ⎟⎟⎟⎜⎜⎜ ⎟⎟⎟
⎪
⎪
⎪
⎜⎜⎜ ⎟⎟⎟⎜⎜⎜
⎜⎜⎝ ⎟⎟⎠⎜⎜⎝
⎟⎟⎟
⎟⎟⎠
∞ r + k − 1 ⎪
⎪
⎪ x n − x
⎪
⎪
⎨ ⎛ ⎞ , x = 0, 1, · · · , n
(1 − x)−r = xk ⎜⎜⎜ M ⎟⎟⎟
f (x) = ⎪
⎪ ⎜⎜⎜ ⎟⎟⎟
k=0 k ⎪
⎪
⎪
⎜⎜⎜
⎝
⎟⎟⎟
⎠
⎪
⎪
⎪ n
Set x = 1 − p and k = x ⎪
⎪
⎪
⎩ 0, otherwise,
 ∞ r + x − 1 ∞ r + x − 1
−r x
p = (1−p) , i.e., 1 = pr (1−p) x where M = 1, 2, · · ·, K = 0, 1, · · · , M, and n = 1, 2, · · · , M.
x=0 x x=0 x

38 39

In Example 1.2 (Section 1.1), all the possible values of X are The probability that X takes x1 , x2 , x3 or x4 is given by:
0, 1, 2 and 3. (note that X denotes the number of heads when 1
P(X = 0) = f (0) = P({ω8 }) = ,
a coin is cast three times). 8
P(X = 1) = f (1) = P({ω4 , ω6 , ω7 })
That is, x1 = 0, x2 = 1, x3 = 2 and x4 = 3 are assigned in
3
= P({ω4 }) + P({ω6 }) + P({ω7 }) = ,
this case. 8
P(X = 2) = f (2) = P({ω2 , ω3 , ω5 })
3
= P({ω2 }) + P({ω3 }) + P({ω5 }) = ,
8
1
P(X = 3) = f (3) = P({ω1 }) = ,
8
which can be written as:

40 41

Continuous Random Variable (連続型確率変数) and Prob-

3! 1
x 1
3−x ability Density Function (確率密度関数): Whereas a dis-
P(X = x) = f (x) = , x = 0, 1, 2, 3.
x!(3 − x)! 2 2 crete random variable assumes at most a countable set of

For P(X = 1) and P(X = 2), note that each sample point is possible values, a continuous random variable X takes any

mutually exclusive. real number within an interval I.

The above probability function is called the binomial distri- For the interval I, the probability which X is contained in A

bution (二項分布). is defined as: 

 P(X ∈ I) = f (x) dx.
Thus, it is easy to check f (x) ≥ 0 and x f (x) = 1 in Exam-
I
ple 1.2.

42 43
For example, let I be the interval between a and b for a < b. In order for f (x) to be a probability density function, f (x)
Then, we can rewrite P(X ∈ I) as follows: has to satisfy the following properties:
 b
P(a < X < b) = f (x) dx, f (x) ≥ 0,
a  ∞
f (x) dx = 1.
where f (x) is called the probability density function (確率 −∞

密度関数) of X, or simply the density function (密度関数)

of X.

44 45

Some functional forms of f (x) are as follows: Normmal distribution (正規分布):

1 1

Uniform distribution (一様分布): f (x) = √ exp − 2 (x − μ)2 , −∞ < x < ∞

⎧ 2πσ2 2σ
⎪
⎪
⎪ 1
⎨ b − a,
⎪ a<x<b
where −∞ < μ < ∞ and σ > 0.
f (x) = ⎪
⎪
⎪
⎪
⎩ 0, otherwise, X ∼ N(μ, σ2 )
where −∞ < a < b < ∞. N(0, 1) = Standard normal distribution
X ∼ U(a, b)

46 47

Exponential distribution (指数分布): Gamma distribution (ガンマ分布):

⎧ r
⎧ ⎪
⎪ λ r−1 −λx
⎪
⎪ −λx ⎪
⎨ Γ(r) x e ,
⎪ x>0
⎪
⎨ λe , x>0
f (x) = ⎪ f (x) = ⎪
⎪
⎪
⎪ ⎪
⎪
⎩ 0,
⎩ 0, otherwise, otherwise,
where λ > 0 and r > 0.
where λ > 0
Gamma dist. with r = 1 ⇐⇒
 ∞ Exponential dist.
Gamma function: Γ(a) = xa−1 e−x dx, a > 0
0
Γ(a+1) = aΓ(a) —> Use integration by parts (部分積分)
Γ(n + 1) = n! for integer n
1 1 · 3 · 5 · · · (2n − 1) √ 1 3 √
Γ(n+ ) = π, Γ( ) = 2Γ( ) = π
2 2n 2 2

48 49
Beta distribution (ベータ分布): Cauchy distribution (コーシー分布):
⎧
⎪
⎪
⎪
1 1
⎪
⎨ xa−1 (1 − x)b−1 , 0<x<1 f (x) = , −∞ < x < ∞
f (x) = ⎪ B(a, b) πβ(1 + (x − α)2 /β2 )
⎪
⎪
⎪
⎩ 0, otherwise,
where −∞ < α < ∞ and β > 0.
where a > 0 and b > 0.
 1 Log-normal distribution (対数正規分布):
Beta function: B(a, b) = xa−1 (1 − x)b−1 dx
⎧ 1

0
⎪
⎪
⎪
1
Γ(a)Γ(b) ⎨ √
⎪ exp − 2 (ln x − μ)2 , 0<x<∞
= f (x) = ⎪
⎪ x 2πσ2 2σ
Γ(a + b) ⎪
⎪
⎩ 0, otherwise,
B(a, b) = B(b, a)
where −∞ < μ < ∞ and σ > 0.

50 51

Double exponentioal distribution (二重指数分布), or Laplace Logistic distribution (ロジスティック分布):

distribution (ラプラス分布):
F(x) = (1 + e−(x−α)/β )−1 , −∞ < x < ∞
1 |x − α|

f (x) = exp − , −∞ < x < ∞

2β β where −∞ < α < ∞ and β > 0.
where −∞ < α < ∞ and β > 0.
Pareto distribution (パレート分布):
Weibull distribution (ワイブル分布): ⎧
⎪
⎪
⎪ θxθ
⎧ ⎪ θ+10 ,
⎨ x > x0
⎪
⎪
⎪ b−1 b f (x) = ⎪ x
⎨ abx exp(−ax ), x>0 ⎪
⎪
⎪
f (x) = ⎪
⎪ ⎩ 0, otherwise,
⎪
⎩ 0, otherwise,
where x0 > 0 and θ > 0.
where a > 0 and b > 0.

52 53

Gumbel distribution (ガンベル分布), or Extreme value t distribution (t 分布):

distribution (極値分布): Γ( k+1 ) 1 x2
−(k+1)/2
f (x) = 2
√ 1 + , −∞ < x < ∞
Γ( 2k ) kπ k
−(x−α)/β
F(x) = exp(−e ), −∞ < x < ∞
where k > 0.
where −∞ < α < ∞ and β > 0. X ∼ t(k) —> t dist. with k degrees of freedom (自由度)
t(1) ⇐⇒ Cauchy dist.
t(∞) ⇐⇒ N(0, 1)

54 55
F distribution (F 分布): Chi-square distribution (カイ二乗分布):
⎧ ⎧
⎪
⎪
⎪ Γ( m+n ) m
m/2 xm/2−1 ⎪
⎪
⎪ 1 k/2−1 −x/2
⎪ , x>0 ⎨ Γ( k )2k/2 x
⎪ e , x>0
2
⎪
⎨ Γ( )Γ( ) n
m n ⎪
f (x) = ⎪ (1 + mn x)(m+n)/2 f (x) = ⎪
⎪
⎪
⎪
2 2 ⎪
⎪
⎪
2
⎪
⎩ 0, otherwise, ⎪
⎩ 0, otherwise,

where m, n = 1, 2, · · ·. where k = 1, 2, · · ·.
X ∼ F(m, n) —> F dist. with (m, n) degrees of freedom X ∼ χ2 (k) —> χ2 dist. with k degrees of freedom

56 57

For a continuous random variable, note as follows:

 x
P(X = x) = f (t) dt = 0.
x

In the case of discrete random variables, P(X = xi ) represents

the probability which X takes xi , i.e., pi = f (xi ).
Thus, the probability function f (xi ) itself implies probability.
However, in the case of continuous random variables, P(a <
X < b) indicates the probability which X lies on the interval
(a, b).

58