Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Exam P Formula Sheet

0 𝑖𝑓 𝑥 < 𝑑
Deductible of d 𝑌 = max(0, 𝑋 − 𝑑) = {
𝑋 − 𝑑 𝑖𝑓 𝑋 ≥ 𝑑

−𝑏±√𝑏2 −4𝑎𝑐
Quadratic Formula 2𝑎

𝑎(1−𝑟 𝑛 )
Geometric Sequence ∑𝑛𝑖=1 𝑎 ∗ 𝑟 𝑖−1 = 1−𝑟

2
∑𝑛 ̅̅̅2
𝑖=1(𝑥𝑖 −𝑥) 𝑛(∑𝑛 2 𝑛
𝑖=1 𝑥𝑖 )−(∑𝑖=1 𝑥𝑖 )
Variance 𝑠2 = 𝑛−1
= 𝑛(𝑛−1)

Standard Deviation 𝑠 = √𝑠 2

Demorgan’s Laws ̅̅̅̅̅̅̅̅̅


𝐴 ∪ 𝐵 = 𝐴̅ ∪ 𝐵̅ , ̅̅̅̅̅̅̅̅̅
𝐴 ∩ 𝐵 = 𝐴̅ ∩ 𝐵̅

Distributive Laws 𝐴 ∩ (𝐵 ∪ 𝐶) = (𝐴 ∩ 𝐵) ∪ (𝐴 ∩ 𝐶), 𝐴 ∪ (𝐵 ∩ 𝐶) = (𝐴 ∪ 𝐵) ∩ (𝐴 ∪ 𝐶)

Probability 𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) − 𝑃(𝐴 ∩ 𝐵)


Mutually exclusive 𝑃(𝐴 ∩ 𝐵) = 0 → 𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵)
Three events 𝑃(𝐴 ∪ 𝐵 ∪ 𝐶) = 𝑃(𝐴) + 𝑃(𝐵) + 𝑃(𝐶) − 𝑃(𝐴 ∩ 𝐵) − 𝑃(𝐴 ∩ 𝐶) − 𝑃(𝐶 ∩ 𝐷) − 𝑃(𝐴 ∩ 𝐵 ∩ 𝐶)

𝑛!
Permutations (r objects from n) 𝑛 𝑃𝑟 = (𝑛−𝑟)!
𝑛! 𝑛 𝑃𝑟
Combination of r object from n total) 𝑛 𝐶𝑟 = 𝑟!(𝑛−𝑟)! = 𝑟!
𝑃(𝐴∩𝐵)
Conditional Probability 𝑃(𝐴 ∩ 𝐵) = 𝑃(𝐴) ∗ 𝑃(𝐵|𝐴) → 𝑃(𝐵|𝐴) = 𝑃(𝐴)
𝑃(𝐵̅|𝐴) = 1 − 𝑃(𝐵|𝐴)

Independence 𝑃(𝐵|𝐴) = 𝑃(𝐴 ∩ 𝐵) when A & B are independent


𝑃(𝐴 ∩ 𝐵) = 𝑃(𝐴) ∗ 𝑃(𝐵) when A & B are independent

𝑃(𝐴𝑖 ∗𝑃(𝐵|𝐴𝑖 )
Bayes’ Rule 𝑃(𝐴𝑖 |𝐵) = ∑𝑁
𝑖=1 𝑃(𝐴𝑖 )∗𝑃(𝐵|𝐴𝑖 )

Discrete Probability f(x) = pmf (probability mass function)


Fx(x) = P (X <= x)

Expected Value 𝐸[𝑔(𝑋)] = ∑𝑥 𝑔(𝑥)𝑓(𝑥)


First Moment 𝑓𝑖𝑟𝑠𝑡 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝜇 = 𝐸(𝑋) = ∑𝑥 𝑥 ∗ 𝑓(𝑥)
Second Moment second 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝐸(𝑋 2 ) = ∑𝑥 𝑥 2 ∗ 𝑓(𝑥)
c = Constant 𝐸[𝑐𝑔(𝑋)] = 𝑐𝐸[𝑔(𝑋)]
Expectation of Sum 𝐸[∑𝑚 𝑚
𝑖=1 𝑔𝑖 (𝑥)] = ∑𝑖=1 𝐸[𝑔𝑖 (𝑥)]
Variance 𝑉(𝑋) = 𝜎𝑥2 = 𝐸[(𝑋 − 𝜇)]2 = 𝐸(𝑋 2 ) − [𝐸(𝑋)]2
𝑉(𝑐) = 0
𝑉[𝑐𝑔(𝑋)] = 𝑐 2 𝑉[𝑔(𝑥)]

Moment Generating Functions 𝑀𝑋 (𝑡) = 𝐸[𝑒 𝑡𝑋 ] = ∑𝑥 𝑒 𝑡𝑥 ∗ 𝑓(𝑥)


𝑑 𝑘 𝑀𝑋 (𝑡)
kth moment 𝑘 𝑡ℎ 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝐸(𝑋 𝑘 ) = 𝑑𝑡 𝑘
|𝑡=0

1
Uniform Distribution 𝑓(𝑥) = 𝑏−𝑎+1 𝑤ℎ𝑒𝑟𝑒 𝑥 = 𝑎, 𝑎 + 1, 𝑎 + 2, … , 𝑏 − 1, 𝑏 𝑎𝑛𝑑 𝑎 ≤ 𝑏
𝑎+𝑏
𝜇𝑋 = 𝐸(𝑋) = 2
(𝑏−𝑎+1)2 −1
𝜎𝑥2 = 𝑉(𝑋) = 12
𝑒 𝑎𝑡 −𝑒 (𝑏+1)𝑡
𝑀𝑋 (𝑡) = (𝑏−𝑎−1)(1−𝑒 𝑡 )

Hypergeometric Distribution – counts number of successes from selecting n objects w/o replacement
from N dichotomous objects
(𝑁1 )( 𝑁2 )
𝑃(𝑋 = 𝑥) = 𝑥 𝑁𝑛−𝑥
(𝑛 )
Where x = 0,1,2,….,n
X <= N1
n – x <=N2
N = N1 + N2

𝑛 𝑛!
( )=
𝑥 𝑥! (𝑛 − 𝑥)!
𝑁1
𝜇𝑋 = 𝐸(𝑋) = 𝑛 ( )
𝑁
2 𝑁1 𝑁 𝑁−𝑛
𝜎𝑥 = 𝑉(𝑋) = 𝑛 ( ) ( 2 ) ( )
𝑁 𝑁 𝑁−1

Binomial Distribution - # of successes in n trials in a two outcome experiment

𝑛!
𝑃(𝑋 = 𝑥) = 𝑝 𝑥 𝑞𝑛−𝑥
𝑥! (𝑛 − 𝑥)!
Where x = 0,1,2,….,n
q = 1 – p, 0 < p < 1
𝜇𝑋 = 𝐸(𝑋) = 𝑛 ∗ 𝑝
𝜎𝑥2 = 𝑉(𝑋) = 𝑛 ∗ 𝑝 ∗ 𝑞
𝑀𝑋 (𝑡) = (𝑞 + 𝑝𝑒 𝑡 )𝑛
Sum of binomial random variables (same p)
Xi ~ Binom(ni,p), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ Binom (n=n1+n2+…+nm, p)
𝑛!
Multinomial Distribution 𝑃(𝑋1 = 𝑥1, 𝑋2 = 𝑥2 , … , 𝑋𝑚 = 𝑥𝑚 ) = 𝑝 𝑥1 𝑝2𝑥2 … 𝑝𝑚
𝑥𝑚
𝑋1 !𝑋2 !…𝑋𝑚 ! 1

Geometric Distribution – counts the number of independent trials need to obtain the first success
𝑃(𝑋 = 𝑥) = 𝑞 𝑥−1 𝑝
Where x = 0,1,2,….,n
q = 1 – p, 0 < p < 1
𝑃(𝑋 > 𝑥) = 𝑞 𝑥 → 𝐹𝑋 (𝑥) = 𝑃(𝑋 ≤ 𝑥) = 1 − 𝑞 𝑥
1
𝜇𝑋 = 𝐸(𝑋) =
𝑝
2
𝑞
𝜎𝑥 = 𝑉(𝑋) = 2
𝑝
𝑝𝑒 𝑡
𝑀𝑋 (𝑡) =
1 − 𝑞𝑒 𝑡
Memoryless property 𝑃(𝑋 > 𝑎 + 𝑏|𝑋 > 𝑎) = 𝑃(𝑋 > 𝑏) , 𝑃(𝑋 < 𝑎 + 𝑏|𝑋 > 𝑎) = 𝑃(𝑋 < 𝑏)

Negative Binomial Distribution – counts the number of trials to obtain the rth success
𝑥 − 1 𝑥−𝑟 𝑟
𝑓(𝑥) = ( )𝑞 𝑝
𝑟−1
Where x= r, r+1, r+2,… and 0 < p < 1
𝑥−1 (𝑥 − 1)!
( )=
𝑟−1 (𝑟 − 1)! (𝑥 − 𝑟)!
𝑟
𝜇𝑋 = 𝐸(𝑋) =
𝑝
𝑟𝑞
𝜎𝑥2 = 𝑉(𝑋) = 2
𝑝
𝑟
𝑝𝑒 𝑡
𝑀𝑋 (𝑡) = ( )
1 − 𝑞𝑒 𝑡
Sum of geometric trials random variables (same p)
Xi ~ GeoT(p), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ NegBinomT (r=m, p)

Sum of negative binomial trials random variables (same p)


Xi ~ NegBinomT(ri,p), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ NegBinomT (r=r1+r2+…+rm, p)

Poisson Distribution – counts the number of events in a fixed interval of time (average =λ)
𝜆𝑥 exp(−𝜆)
𝑃(𝑋 = 𝑥) = 𝑤ℎ𝑒𝑟𝑒 𝑥 = 0,1,2, … 𝑎𝑛𝑑 𝜆 > 0
𝑥!
𝜇𝑋 = 𝐸(𝑋) = 𝜆
𝜎𝑥2 = 𝑉(𝑋) = 𝜆
𝑀𝑋 (𝑡) = exp[ 𝜆(𝑒 𝑡 − 1)]
Sum of Poisson random variables (independent)
Xi ~ Pois(𝜆𝑖), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ Pois (𝜆 = 𝜆1 + 𝜆2 + ⋯ + 𝜆𝑚)

Geometric Failures Distribution – number of failures to obtain first success


𝑃(𝑌 = 𝑦) = 𝑞 𝑦 𝑝
Where y = 0,1,2,…
q = 1 – p, 0 < p < 1
𝑞
𝜇𝑌 = 𝐸(𝑌) =
𝑝
2
𝑞
𝜎𝑌 = 𝑉(𝑌) = 2
𝑝
𝑝
𝑀𝑌 (𝑡) =
1 − 𝑞𝑒 𝑡
Sum of geometric failures random variables (same p)
Xi ~ GeoF(p), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ NegBinomF (r=m, p)

Negative Binomial Failures Distribution – number of failures to obtain the rth success
𝑦+𝑟−1 𝑦 𝑟
𝑃(𝑌 = 𝑦) = ( )𝑞 𝑝
𝑟−1
Y=0,1,2,… and 0 < p < 1 and q=1-p
𝑦+𝑟−1 (𝑦 + 𝑟 − 1)!
( )=
𝑟−1 𝑦! (𝑟 − 1)!
𝑟𝑞
𝜇𝑌 = 𝐸(𝑌) =
𝑝
𝑟𝑞
𝜎𝑥2 = 𝑉(𝑋) = 2
𝑝
𝑝 𝑟
𝑀𝑌 (𝑡) = ( )
1 − 𝑞𝑒 𝑡
Sum of negative binomial failures random variables (same p)
Xi ~ NegBinomF(ri,p), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ NegBinomF (r=r1+r2+…+rm, p)


Continuous Probability ∫−∞ 𝑓(𝑥) = 1 , 𝑓(𝑥) ≥ 0
𝑏
𝑃(𝑎 ≤ 𝑋 ≤ 𝑏) = ∫ 𝑓(𝑥)𝑑𝑥
𝑎
𝑃(𝑎 ≤ 𝑋 ≤ 𝑏) = 𝑃(𝑋 ≤ 𝑏) − 𝑃(𝑋 ≤ 𝑎) = 𝑃(𝑋 ≥ 𝑎) − 𝑃(𝑋 ≥ 𝑏)
𝑥
𝐹𝑋 (𝑥) = 𝑃(𝑋 ≤ 𝑥) = ∫ 𝑓(𝑡)𝑑𝑡 → 𝐹𝑋 (−∞) = 0, 𝐹𝑋 (∞) = 1
−∞
𝑑
𝑓(𝑥) = 𝐹 (𝑥)
𝑑𝑥 𝑋

Expected Value 𝐸[𝑔(𝑋)] = ∫−∞ 𝑔(𝑥)𝑓(𝑥) 𝑑𝑥

First Moment 𝑓𝑖𝑟𝑠𝑡 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝜇 = 𝐸(𝑋) = ∫−∞ 𝑥 ∗ 𝑓(𝑥) 𝑑𝑥

Second Moment second 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝐸(𝑋 2 ) = ∫−∞ 𝑥 2 ∗ 𝑓(𝑥)𝑑𝑥
c = Constant 𝐸[𝑐𝑔(𝑋)] = 𝑐𝐸[𝑔(𝑋)]
Expectation of Sum 𝐸[∑𝑚 𝑚
𝑖=1 𝑔𝑖 (𝑥)] = ∑𝑖=1 𝐸[𝑔𝑖 (𝑥)]

Variance 𝑉(𝑋) = 𝜎𝑥2 = 𝐸[(𝑋 − 𝜇)]2 = 𝐸(𝑋 2 ) − [𝐸(𝑋)]2


𝑉(𝑐) = 0
𝑉[𝑐𝑔(𝑋)] = 𝑐 2 𝑉[𝑔(𝑥)]


Moment Generating Functions 𝑀𝑋 (𝑡) = 𝐸[𝑒 𝑡𝑋 ] = ∫−∞ 𝑒 𝑡𝑥 ∗ 𝑓(𝑥) 𝑑𝑥
𝑑 𝑘 𝑀𝑋 (𝑡)
kth moment 𝑘 𝑡ℎ 𝑚𝑜𝑚𝑒𝑛𝑡 = 𝐸(𝑋 𝑘 ) = 𝑑𝑡 𝑘
|𝑡=0

1
Uniform Distribution 𝑓(𝑥) = 𝑏−𝑎 𝑤ℎ𝑒𝑟𝑒 𝑎 ≤ 𝑥 ≤ 𝑏
𝑥−𝑎
𝐹(𝑋) = 𝑃(𝑋 ≤ 𝑥) =
𝑏−𝑎
𝑎+𝑏
𝜇𝑋 = 𝐸(𝑋) =
2
2 (𝑏−𝑎)2
𝜎𝑥 = 𝑉(𝑋) = 12
𝑒 𝑡𝑏 −𝑒 𝑡𝑎
𝑀𝑋 (𝑡) =
𝑡(𝑏−𝑎)

Exponential Distribution – waiting time for the first observation of an event


1
𝑓(𝑥) = exp(−𝑥⁄𝜃 ) 𝑤ℎ𝑒𝑟𝑒 𝑥 > 0, 𝜃 > 0
𝜃
𝑃(𝑎 < 𝑥 < 𝑏 = exp(−𝑎⁄𝜃 ) − exp(−𝑏⁄𝜃 )
𝑥=𝑏
1
∫ 𝑥 exp(−𝑥⁄𝜃 )𝑑𝑥 = exp(−𝑎⁄𝜃 ) (𝜃 + 𝑎) − exp(−𝑏⁄𝜃 )(𝜃 + 𝑏)
𝑥=𝑎 𝜃
𝐹(𝑥) = 𝑃(𝑋 ≤ 𝑥) = 1 − exp(−𝑥⁄𝜃 )
𝑃(𝑋 > 𝑥) = exp(−𝑥⁄𝜃 )
𝜇𝑋 = 𝐸(𝑋) = 𝜃
𝜎𝑥2 = 𝑉(𝑋) = 𝜃 2
𝑀𝑋 (𝑡) = (1 − 𝜃𝑡)−1
Log properties
ln(𝑎𝑏 ) = 𝑏 ∗ ln(𝑎)
ln(𝑒 𝑎 ) = 𝑒 ln(𝑎) = 𝑎
𝑎
ln(𝑎 ∗ 𝑏) = ln(𝑎) + ln(𝑏) → ln ( ) = ln(𝑎) − ln(𝑏
𝑏
Memoryless property 𝑃(𝑋 > 𝑎 + 𝑏|𝑋 > 𝑎) = 𝑃(𝑋 > 𝑏) , 𝑃(𝑋 < 𝑎 + 𝑏|𝑋 > 𝑎) = 𝑃(𝑋 < 𝑏)
Similar to Poisson distribution – waiting time for event
Gamma Distribution
1
𝑓(𝑥) = 𝑥 𝛼−1 exp(−𝑥⁄𝜃 ) 𝑤ℎ𝑒𝑟𝑒 𝑥 > 0, 𝜃 > 0, 𝛼 > 0
𝜃 𝛼 (𝛼
− 1)!
1 𝑏 −𝑥
𝛼 = 1, ∫ exp(−𝑥⁄𝜃 ) 𝑑𝑥 = −𝑒 ⁄𝜃 |𝑏𝑎
𝜃 𝑎
1 𝑏 −𝑥 𝑥
𝛼 = 2, 2 ∫ x ∗ exp(−𝑥⁄𝜃 ) 𝑑𝑥 = −𝑒 ⁄𝜃 (1 + ) |𝑏𝑎
𝜃 𝑎 𝜃
𝑏
1 −𝑥 𝑥 𝑥2
𝛼 = 3, 3 ∫ x 2 ∗ exp(−𝑥⁄𝜃 ) 𝑑𝑥 = −𝑒 ⁄𝜃 (1 + + 2 ) |𝑏𝑎
2𝜃 𝑎 𝜃 2𝜃
𝜇𝑋 = 𝐸(𝑋) = 𝛼 ∗ 𝜃
𝜎𝑥2 = 𝑉(𝑋) = 𝛼 ∗ 𝜃 2
𝑀𝑋 (𝑡) = (1 − 𝜃𝑡)−𝛼
Sum of exponential random variables (same 𝜃)
Xi ~ Exp(𝜃), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ Gamma (𝛼 =m, 𝜃)

Sum of gamma random variables (same 𝜃)


Xi ~ Gamma(𝛼𝑖, 𝜃), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ Gamma (𝛼 = 𝛼1 + 𝛼2 + ⋯ + 𝛼𝑚, 𝜃)

Normal Distribution
1 1 𝑥−𝜇 2
𝑓(𝑥) = 𝑒𝑥𝑝 [− ( ) ]
√2𝜋𝜎 2 𝜎
−∞ < 𝑥 < ∞, −∞ < 𝜇 < ∞, 0 < 𝜎 < ∞
𝐸(𝑋) = 𝜇, 𝑉(𝑋) = 𝜎 2
𝑋 − 𝜇𝑋
𝑍=
𝜎𝑋
1
𝑀𝑋(𝑡) = 𝑒𝑥𝑝 (𝜇𝑡 + 𝜎 2 𝑡 2 )
2
Sum of normal random variables
Xi ~ N(µi, 𝜎 2 𝑖), i = 1, 2, …, m
Let X = X1 + X2 + …+ Xm. Then
X ~ N (µ= µ1+ µ2+…+ µm, 𝜎 2 = 𝜎 2 1 + 𝜎 2 2 + ⋯ + 𝜎 2 𝑚)

Continuous Transformations
Method of Distributions
1. Let Y=f(X). Solve for X (𝑋 = 𝑓 −1 (𝑌)
2. Calculate cdf of Y:𝐹𝑋 (𝑌) = 𝑃(𝑌 ≤ 𝑦) = 𝑃(𝑓(𝑋) ≤ 𝑦) = 𝑃(𝑋 ≤ 𝑓 −1 (𝑦)) = 𝐹𝑋 (𝑓 −1 (𝑦))
3. Differentiate answer from #2 to obtain f(y)
4. f(y) and Fy(Y) are now know.
Method of Transformations
1. Let Y=f(X). Solve for X (𝑋 = 𝑓 −1 (𝑌)
𝑑𝑋 𝑑
2. Differentiate answer from #1 to find 𝑑𝑌 = 𝑑𝑌 𝑓 −1 (𝑌)
𝑑𝑋 𝑑𝑋
3. 𝑓𝑌 (𝑦) = 𝑓𝑋 (𝑥) | | = 𝑓𝑋 (𝑓 −1 (𝑦)) | |
𝑑𝑌 𝑑𝑌
4. f(y) and Fy(Y) are now know.

Order Statistics
Minimum of N identically distributed variables
𝐹𝑋(1) (𝑥) = 1 − [1 − 𝐹𝑥 (𝑥)]𝑁
𝑓𝑋(1) (𝑥) = 𝑁[1 − 𝐹𝑋 (𝑥)]𝑁−1 𝑓(𝑥)
Maximum of N identically distributed variables
𝐹𝑋(𝑁) (𝑥) = [𝐹𝑥 (𝑥)]𝑁
𝑓𝑋(𝑁) (𝑥) = 𝑁[𝐹𝑋 (𝑥)]𝑁−1 𝑓(𝑥)

Multivariate Distributions
Discrete ∑𝑥 ∑𝑦 𝑃(𝑋 = 𝑥, 𝑌 = 𝑦) = 1 , 𝑃(𝑋 = 𝑥, 𝑌 = 𝑦) ≥ 0
Continuous ∬ 𝑓(𝑥, 𝑦)𝑑𝑦𝑑𝑥 = 1, 𝑓(𝑥, 𝑦) ≥ 0

Marginal of X: 𝑃(𝑋 = 𝑥) = ∑𝑦 𝑃(𝑋 = 𝑥, 𝑌 = 𝑦)


𝑓𝑋 (𝑥) = ∫𝑦 𝑓(𝑥, 𝑦)𝑑𝑦
Marginal of Y: 𝑃(𝑌 = 𝑦) = ∑𝑋 𝑃(𝑋 = 𝑥, 𝑌 = 𝑦)
𝑓𝑌 (𝑦) = ∫𝑥 𝑓(𝑥, 𝑦)𝑑𝑥
𝑃(𝑋,𝑌) 𝑓(𝑥.𝑦)
Conditional of X given Y: 𝑃(𝑋|𝑌) = 𝑃(𝑌)
= 𝑓(𝑦)
𝑃(𝑋,𝑌) 𝑓(𝑥.𝑦)
Conditional of Y given X: 𝑃(𝑌|𝑋) = =
𝑃(𝑋) 𝑓(𝑥)
Therefore: 𝑓(𝑥, 𝑦) = 𝑓𝑌 (𝑦)𝑓𝑋|𝑌 (𝑥|𝑦) = 𝑓𝑋 (𝑥)𝑓𝑌|𝑋 (𝑦|𝑥)
Discrete cdf: 𝐹𝑋,𝑌 (𝑥, 𝑦) = 𝑃(𝑋 ≤ 𝑥, 𝑌 ≤ 𝑦)
𝑥 𝑦
Continuous cdf: ∫−∞ ∫−∞ 𝑓(𝑥, 𝑦)𝑑𝑦𝑑𝑥
X and Y are independent iff 𝑓𝑋,𝑌 (𝑥, 𝑦) = 𝑓𝑋 (𝑥) ∗ 𝑓𝑌 (𝑦)

First conditional moment 𝐸(𝑋|𝑌) = 𝜇𝑋|𝑌 = ∫−∞ 𝑥 ∗ 𝑓(𝑥|𝑦)𝑑𝑥
𝐸(𝑋|𝑌) = 𝜇𝑋|𝑌 = ∑ 𝑥 ∗ 𝑓(𝑥|𝑦)
𝑥

Second conditional moment 𝐸(𝑋 2 |𝑌) = ∫−∞ 𝑥 2 ∗ 𝑓(𝑥|𝑦)𝑑𝑥
𝐸(𝑋 2 |𝑌) = ∑ 𝑥 2 ∗ 𝑓(𝑥|𝑦)
𝑥

Conditional Expectations 𝐸(𝑔(𝑋|𝑌)) = ∫−∞ 𝑔(𝑥)𝑓(𝑥|𝑦)𝑑𝑥
𝐸(𝑔(𝑋|𝑌)) = ∑ 𝑔(𝑥)𝑓(𝑥|𝑦)
𝑥
2
Variance 𝑉(𝑋|𝑌) = 𝜎𝑋|𝑌 = 𝐸(𝑋 2 |𝑌) − [𝐸(𝑋|𝑌)]2
Expected Value 𝐸(𝑋) = 𝐸𝑌 [𝐸𝑋 (𝑋|𝑌)]
Variance of X 𝑉(𝑋) = 𝑉𝑌 [𝐸𝑌 (𝑋|𝑌)] + 𝐸𝑌 [𝑉𝑋 (𝑋|𝑌)]

Expectation Properties
𝐸[𝑔(𝑥, 𝑦)] = ∑ ∑ 𝑔(𝑥, 𝑦)𝑓(𝑥, 𝑦)
𝑥 𝑦

𝐸[𝑔(𝑥, 𝑦)] = ∬ 𝑔(𝑥, 𝑦)𝑓(𝑥, 𝑦)𝑑𝑦𝑑𝑥


Covariance 𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸[(𝑋 − 𝜇𝑋 )(𝑌 − 𝜇𝑌 )] = 𝐸(𝑋𝑌) − 𝐸(𝑋)𝐸(𝑌)
𝐶𝑜𝑣(𝑋, 𝑌) = 𝐶𝑜𝑣(𝑌, 𝑋)
𝐶𝑜𝑣(𝑐, 𝑋) = 0
𝐶𝑜𝑣(𝑐𝑋, 𝑑𝑌) = 𝑐𝑑𝐶𝑜𝑣(𝑋, 𝑌)
Independence 𝐸[𝑔(𝑥)ℎ(𝑦)] = 𝐸[𝑔(𝑥)]𝐸[ℎ(𝑦)] 𝑤ℎ𝑒𝑛 𝑋 𝑎𝑛𝑑 𝑌 𝑎𝑟𝑒 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡
𝐶𝑜𝑣(𝑋, 𝑌) = 0 𝑤ℎ𝑒𝑛 𝑋 𝑎𝑛𝑑 𝑌 𝑎𝑟𝑒 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡
𝐶𝑜𝑣(𝑋,𝑌)
Correlation coefficient 𝜌𝑋,𝑌 = 𝜎𝑋 𝜎𝑌
Expectation of a sum is the sum of expectations 𝐸(∑𝑚 𝑚
𝑖=1 𝑎𝑖 𝑋𝑖 ) = ∑𝑖=1 𝑎𝑖 𝐸(𝑋𝑖 )
Variance of a sum 𝑉(∑𝑚 𝑚 2
𝑖=1 𝑎𝑖 𝑋𝑖 ) = ∑𝑖=1 𝑎𝑖 𝑉(𝑋𝑖 ) + 2 ∑𝑖<𝑗 ∑ 𝑎𝑖 𝑎𝑗 𝐶𝑜𝑣(𝑋𝑖 , 𝑌𝑖 )
For m=2 𝑉(𝑎1 𝑋1 + 𝑎2 𝑋2 ) = 𝑎12 𝑉(𝑋1 ) + 𝑎22 𝑉(𝑋2 ) + 2𝑎1 𝑎2 𝐶𝑜𝑣(𝑋1 , 𝑋2 )
When independent (∑𝑚 𝑚 2
𝑖=1 𝑎𝑖 𝑋𝑖 ) = ∑𝑖=1 𝑎𝑖 𝑉(𝑋𝑖 )
Covariance of sums is the sum of covariances
𝐶𝑜𝑣(∑𝑚 𝑛 𝑚 𝑛
𝑖=1 𝑎𝑖 𝑋𝑖 , ∑𝑗=1 𝑏𝑗 𝑌𝑗 , ) = ∑𝑖=1 ∑𝑗=1 𝑎𝑖 𝑏𝑗 𝐶𝑜𝑣(𝑋𝑖 , 𝑌𝑗 )
Sum of a random number of random variables
𝑁

𝑆 = ∑ 𝑐𝑋𝑖 = 𝑐𝑋1 + 𝑐𝑋2 + ⋯ + 𝑐𝑋𝑁


𝑖=1
𝐸(𝑆) = 𝑐𝐸(𝑋)𝐸(𝑁) 𝑖𝑓 𝑋𝑖 𝑎𝑟𝑒 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑎𝑛𝑑 𝑖𝑑𝑒𝑛𝑡𝑖𝑐𝑎𝑙𝑙𝑦 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑒𝑑
𝑉(𝑆) = 𝑐 2 {[𝐸(𝑋)]2 𝑉(𝑁) + 𝑉(𝑋)𝐸(𝑁)𝑖𝑓 𝑋𝑖 𝑎𝑟𝑒 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑎𝑛𝑑 𝑖𝑑𝑒𝑛𝑡𝑖𝑐𝑎𝑙𝑙𝑦 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑒𝑑
Joint moment generating functions
𝑀𝑋,𝑌 (𝑡, 𝑠) = 𝐸[𝑒 𝑡𝑋+𝑠𝑌 ] = ∑ ∑ 𝑒 𝑡𝑋+𝑠𝑌 𝑓(𝑥, 𝑦)
𝑥 𝑦
𝑡𝑋+𝑠𝑌 ] 𝑡𝑋+𝑠𝑌
𝑀𝑋,𝑌 (𝑡, 𝑠) = 𝐸[𝑒 = ∬𝑒 𝑓(𝑥, 𝑦)𝑑𝑦𝑑𝑥
𝑑𝑚 𝑑𝑛
𝐸(𝑋 𝑚 , 𝑌 𝑛 ) = 𝑀 (𝑡, 𝑥)|𝑡,𝑠=0
𝑑𝑡 𝑚 𝑑𝑠 𝑛 𝑋,𝑌
𝑀𝑋 (𝑡) = 𝑀𝑋,𝑌 (𝑡, 𝑠 = 0)
𝑀𝑌 (𝑠) = 𝑀𝑋,𝑌 (𝑡 = 0, 𝑠)
𝑛

If 𝑈 = 𝑖=1 𝑋𝑖 = 𝑋1 + 𝑋2 + ⋯ + 𝑋𝑛 where n is a constant all Xi are independent of all Xj
𝑛

𝑀𝑈 (𝑡) = 𝑀∑𝑛𝑖=1 𝑋𝑖(𝑡) = ∏ 𝑀𝑋𝑖 (𝑡)


𝑖=1

Central Limit Theorem (independent random variables)


𝑛
𝑆=∑ 𝑋𝑖 = 𝑋1 + 𝑋2 + ⋯ + 𝑋𝑛
𝑖=1
𝑆
𝑋̅ =
𝑛
𝐸(𝑆)
𝜇𝑆 = 𝐸(𝑆) = ∑𝑛𝑖=1 𝐸(𝑋𝑖 ) 𝜇𝑋̅ = 𝑛
𝑉(𝑆)
𝜎𝑠2 =V(S)= ∑𝑛𝑖=1 𝑉(𝑋𝑖 ) 𝜎𝑋2̅ = 𝑛2
For identical distributions
𝜇𝑆 = 𝐸(𝑆) = 𝑛𝐸(𝑋) = 𝑛𝜇𝑋 𝜇𝑋̅ = 𝐸(𝑋̅) = 𝐸(𝑋) = 𝜇𝑋
2
𝜎𝑋
𝜎𝑠2 =V(S)= 𝑛(𝑉(𝑋) = 𝑛𝜎𝑋2 𝜎𝑋2̅ = 𝑛

If n is sufficiently large (n>30) 𝑆~𝑁(𝜇𝑆 , 𝜎𝑆 ), 𝑋̅~𝑁(𝜇𝑋̅ , 𝜎𝑋̅ )

𝑋
If 𝑋~𝐵𝑖𝑛𝑜𝑚(𝑛, 𝑝)𝑎𝑛𝑑 𝑖𝑓 𝑝̂ = , 𝑡ℎ𝑒𝑛
𝑛,
𝜇𝑝̂ = 𝐸(𝑝̂ ) = 𝑝
𝑝𝑞
𝜎𝑝2̂ = 𝑉(𝑝̂ ) =
𝑛
𝑋
𝐼𝑓 𝑋~𝐵𝑖𝑛𝑜𝑚(𝑛, 𝑝)𝑎𝑛𝑑 𝑖𝑓 𝑝̂ = 𝑎𝑛𝑑 𝑖𝑓 𝑛 𝑖𝑠 𝑠𝑢𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡𝑙𝑦 𝑙𝑎𝑟𝑔𝑒, 𝑡ℎ𝑒𝑛
𝑛
𝑝𝑞
𝑝̂ ~𝑁(𝜇𝑝̂ = 𝑝, 𝜎𝑝̂ = √
𝑛

Normal Approximation of Discrete Probabilities


Problem Continuity correction
P(X<c) P(X< c-0.5)
P(X<=c) P(X<=c+0.5)
P(X>c) P(X>c+0.5)
P(X>=c) P(X>c-0.5)
P(X=c) P(c-0.5<=x<=c+0.5)

You might also like