Winter 2024 Final Exam

ECE286H1 S
Duration: 150 Minutes

Distribution PMF / PDF Expectation Variance

Binomial 𝑛 𝜇 = 𝑛𝑝 𝜎 2 = 𝑛𝑝(1 − 𝑝)
𝑏(𝑥; 𝑛, 𝑝) = ( ) 𝑝 𝑥 (1 − 𝑝)𝑛−𝑥
𝑥
Multinomial 𝑛 𝑥 𝑥 𝜇𝑖 = 𝑛𝑝𝑖 𝜎𝑖2 = 𝑛𝑝𝑖 (1 − 𝑝𝑖 )
𝑓(𝑥1 , … , 𝑥𝑚 ; 𝑝1 , … , 𝑝𝑚 , 𝑛) = (𝑥 , … , 𝑥 ) 𝑝1 1 … 𝑝𝑚𝑚
1 𝑚
Hypergeometric 𝑘 𝑁−𝑘
( )( ) 𝜇=
𝑛𝑘
𝜎2 =
𝑛𝑘(𝑁−𝑛) 𝑘
(1 − )
𝑥 𝑛−𝑥
ℎ(𝑥; 𝑁, 𝑛, 𝑘) = 𝑁
𝑁 𝑁(𝑁−1) 𝑁
( )
𝑛
Negative Binomial 𝑥 − 1 𝑘 𝑥−𝑘 𝜇=
𝑘
𝜎2 =
𝑘(1−𝑝)
𝑏 ∗ (𝑥; 𝑘, 𝑝) = ( )𝑝 𝑞 ,𝑥 ≥ 𝑘 𝑝 𝑝2
𝑘−1
Geometric 𝑔(𝑥; 𝑝) = 𝑝(1 − 𝑝) 𝑥−1 , 𝑥 ≥ 1 𝜇=
1
𝜎 =2 1−𝑝
𝑝 𝑝2
2
Poisson 𝑝(𝑥; 𝜆𝑡) =
𝑒 −𝜆𝑡 (𝜆𝑡)𝑥 𝜇 = 𝜆𝑡 𝜎 = 𝜆𝑡
𝑥!
Uniform 1 𝐴+𝐵 (𝐵−𝐴)2
𝑓(𝑥; 𝐴, 𝐵) = ,𝐴 ≤ 𝑥 ≤ 𝐵 𝜇= 𝜎2 =
𝐵−𝐴 2 12
Normal (𝑥−𝜇)2 𝜇 𝜎2
1 −
𝑛(𝑥; 𝜇, 𝜎) = 𝑒 2𝜎2 , −∞ < 𝑥 < ∞
√2𝜋𝜎
Standard Normal 𝑥 𝑡2 𝜇=0 𝜎2 = 1
1 −
CDF Φ(𝑥) = ∫ 𝑒 2 𝑑𝑡
√2𝜋 −∞
𝑃(𝐴 ≤ 𝑋 ≤ 𝐵) = Φ(𝐵) − Φ(𝐴)
𝑥
Gamma Distribution 1 − 𝜇 = 𝛼𝛽 𝜎 2 = 𝛼𝛽 2
𝑓(𝑥; 𝛼, 𝛽) = 𝑥 𝛼−1 𝑒 𝛽 ,𝑥 > 0
𝛽 𝛼 Γ(𝛼)
𝑥
Exponential 1 − 1 𝜇=𝛽 𝜎 2 = 𝛽2
𝑓(𝑥; 𝛼 = 1, 𝛽) = 𝑓(𝑥; 𝛽) = 𝑒 𝛽 , 𝑥 ≥ 0, 𝛽 =
Distribution 𝛽 𝜆
𝜈 𝑥
Chi-Squared 𝜈
𝑓 (𝑥; 𝛼 = , 𝛽 = 2) = 𝑓(𝑥; 𝜈) = 𝜈
1
𝑥 2−1 𝑒 −2 , 𝑥 > 0 𝜇=𝜈 𝜎 2 = 2𝜈
Distribution 2 𝜈
22 Γ( )
2

Operations with Sets CDF Properties

𝑥
𝐴 ∩ ∅ = ∅, 𝐴 ∪ ∅ = 𝐴, 𝐴 ∩ 𝐴′ = ∅, 𝐴 ∪ 𝐴′ = 𝑆, 𝑆 ′ = ∅, 𝑃(𝑋 ≤ 𝑥) = 𝐹(𝑥) = ∑𝑡≤𝑥 𝑓(𝑡) = ∫−∞ 𝑓(𝑡) 𝑑𝑡, 𝑥 ∈ ℝ
∅′ = 𝑆, (𝐴′ )′ = 𝐴, (𝐴 ∩ 𝐵)′ = 𝐴′ ∪ 𝐵′, (𝐴 ∪ 𝐵)′ = 𝐴′ ∩ 𝐵′ 𝑃(𝑎 < 𝑋 < 𝑏) = 𝐹(𝑏) − 𝐹(𝑎)
Permutations and Combinations Joint Distribution Properties and Marginals
𝑛! 𝑛! 𝑛 𝑛!
𝑛𝑃𝑟 = (𝑛−𝑟)! , 𝑛𝐶𝑟 = ,( )= 𝑓(𝑥, 𝑦) ≥ 0 ∀(𝑥, 𝑦) ∈ 𝑆
𝑟!(𝑛−𝑟)! 𝑛1 , … , 𝑛𝑚 𝑛1 !𝑛2 !…𝑛𝑚! ∞ ∞
Additive and Product Rules ∑𝑥 ∑𝑦 𝑓(𝑥, 𝑦) = ∫−∞ ∫−∞ 𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦 = 1
𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) − 𝑃(𝐴 ∩ 𝐵) 𝑃((𝑋, 𝑌) ∈ 𝐴) = ∑(𝑥,𝑦)∈𝐴 𝑓(𝑥, 𝑦) = ∫(𝑥,𝑦)∈𝐴 𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦
∞
𝑃(𝐴 ∩ 𝐵) = 𝑃(𝐴)𝑃(𝐵|𝐴), 𝑃(𝐴) > 0 𝑔(𝑥) = ∑𝑦 𝑓(𝑥, 𝑦) = ∫−∞ 𝑓(𝑥, 𝑦)𝑑𝑦
Independence, Conditional Probability, and Bayes’ Rule ∞
ℎ(𝑦) = ∑𝑥 𝑓(𝑥, 𝑦) = ∫−∞ 𝑓(𝑥, 𝑦)𝑑𝑥
𝑃(𝐴|𝐵) = 𝑃(𝐴) or 𝑃(𝐵|𝐴) = 𝑃(𝐵) 𝑃(𝐴 ∩ 𝐵) = 𝑃(𝐴)𝑃(𝐵)
𝑃(𝐴∩𝐵) 𝑃(𝐴|𝐵)𝑃(𝐵) Conditional Distributions and Independence of RVs
𝑃(𝐵|𝐴) = , 𝑃(𝐵|𝐴) = , 𝑃(𝐴) > 0 𝑓(𝑥,𝑦)
𝑃(𝐴) 𝑃(𝐴) 𝑓(𝑥|𝑦) =
𝑔(𝑥)
Total Probability Theorem and Corresponding Bayes’ Rule
𝑃(𝑎 ≤ 𝑋 ≤ 𝑏|𝑌 = 𝑐) = ∑𝑎≤𝑥≤𝑏 𝑓(𝑥|𝑌 = 𝑐) =
𝑃(𝐴) = ∑𝑘𝑖=1 𝑃(𝐴 ∩ 𝐵𝑖 ) for partition 𝐵1 , … , 𝐵𝑘
∫𝑎≤𝑥≤𝑏 𝑓(𝑥|𝑌 = 𝑐)𝑑𝑥
𝑃(𝐵)𝑃(𝐴|𝐵)
𝑃(𝐵|𝐴) = ∑𝑘 for partition 𝐶1 , … , 𝐶𝑘 𝑓(𝑥, 𝑦) = 𝑔(𝑥)ℎ(𝑦) (independence)
𝑖=1 𝑃(𝐶𝑖 )𝑃(𝐴|𝐶𝑖 )
PMF and PDF Properties Expectations of Functions of RVs (One and Two)
∞
𝑓(𝑥) ≥ 0, for each outcome/value 𝑋 = 𝑥 𝐸[𝑧(𝑋)] = ∑𝑥 𝑧(𝑥)𝑓(𝑥) = ∫−∞ 𝑧(𝑥)𝑓(𝑥)𝑑𝑥
∞
∑𝑥 𝑓(𝑥) = ∫−∞ 𝑓(𝑥)𝑑𝑥 = 1 normalization 𝐸[𝑧(𝑋, 𝑌)] =
∞ ∞
𝑃(𝑋 = 𝑥) = 𝑓(𝑥) (discrete), 𝑃(𝑎 < 𝑋 < 𝑏) =
𝑏 ∑𝑥 ∑𝑦 𝑧(𝑥, 𝑦)𝑓(𝑥, 𝑦) = ∫−∞ ∫−∞ 𝑧(𝑥, 𝑦)𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦
∫𝑎 𝑓(𝑥) 𝑑𝑥
(continuous) 𝐸[𝑎𝑋 + 𝑏𝑌] = 𝑎𝐸[𝑋] + 𝑏𝐸[𝑌] (any case), 𝐸[𝑋𝑌] =
𝐸[𝑋]𝐸[𝑌] (X and Y independent only)
Winter 2024 Final Exam
ECE286H1 S
Duration: 150 Minutes

Variance Properties Moments and Moment-Generating Functions

𝜎 2 = 𝑉𝑎𝑟(𝑋) = 𝐸[(𝑋 − 𝜇)2 ] = 𝐸[𝑋 2 ] − 𝜇 2 𝑑 𝑟 𝑀𝑋 (𝑡) th
𝜇𝑟′ = 𝐸[𝑋 𝑟 ] = |𝑡=0 , r moment about origin of X
∞ 𝑑𝑡 𝑟
𝜎 2 = ∑𝑥(𝑥 − 𝜇)2 𝑓(𝑥) = ∫−∞(𝑥 − 𝜇)2 𝑓(𝑥)𝑑𝑥 ∞
𝜇𝑟′ = ∑𝑥 𝑟 𝑓(𝑥) = ∫−∞ 𝑥 𝑟 𝑓(𝑥) 𝑑𝑥
2
𝜎𝑎𝑋+𝑏𝑌+𝑐 = 𝑎2 𝜎𝑋2 + 𝑏 2 𝜎𝑌2 + 2𝑎𝑏𝜎𝑋𝑌 (any case)
𝜇 = 𝜇1′ , 𝜎 2 = 𝜇2′ − 𝜇 2
Covariance Properties ∞
𝑀𝑋 (𝑡) = 𝐸[𝑒 𝑡𝑋 ] = ∑𝑒 𝑡𝑥 𝑓(𝑥) = ∫−∞ 𝑒 𝑡𝑥 𝑓(𝑥) 𝑑𝑥 , MGF
𝜎𝑋𝑌 = 𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸[(𝑋 − 𝜇𝑋 )(𝑌 − 𝜇𝑌 )] = 𝐸[𝑋𝑌] − 𝜇𝑋 𝜇𝑌
𝑡2 𝜎2
𝜇𝑡+( )
𝜎𝑋𝑌 = ∑𝑥 ∑𝑦(𝑥 − 𝜇𝑥 )(𝑦 − 𝜇𝑦 )𝑓(𝑥, 𝑦)(discrete) 𝑀𝑋 (𝑡) = 𝑒 2 (Normal Distribution)
∞ ∞
𝜎𝑋𝑌 = ∫−∞ ∫−∞(𝑥 − 𝜇𝑋 )(𝑦 − 𝜇𝑌 )𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦 (continuous) Linear Combinations of RVs
𝜎𝑋𝑌 = 𝐸[𝑋𝑌] − 𝐸[𝑋]𝐸[𝑌] ∴ 𝜎𝑋𝑌 = 0 for X and Y independent For 𝑌 = 𝑎𝑋 given distribution 𝑓(𝑥)
1 𝑦
Correlation Coefficient ℎ(𝑦) = |𝑎| 𝑓 ( )
𝑎
𝜎𝑋𝑌
𝜌𝑋𝑌 = , −1 ≤ 𝜌𝑋𝑌 ≤ 1 𝑀𝑎𝑋+𝑏𝑌+𝑐 (𝑡) = 𝑀𝑋 (𝑎𝑡)𝑀𝑌 (𝑏𝑡)𝑒 𝑐𝑡
𝜎𝑋 𝜎𝑌
Poisson Approximation for Binomial For 𝑍 = 𝑋 + 𝑌 and distributions 𝑓(𝑥), 𝑔(𝑦)
𝑏(𝑥; 𝑛, 𝑝) → 𝑝(𝑥; 𝜆𝑡) as 𝑛 → ∞, 𝑝 → 0, 𝑛𝑝 constant X and Y independent, 𝑋 = 𝑊, 𝑌 = 𝑍 − 𝑊
∞
Chebyshev’s Theorem (Discrete and Continuous RV) ℎ(𝑧) = ∑∞−∞ 𝑓(𝑤)𝑔(𝑧 − 𝑤) = ∫−∞ 𝑓(𝑤)𝑔(𝑧 − 𝑤) 𝑑𝑤
1
𝑃(𝜇 − 𝑘𝜎 < 𝑋 < 𝜇 + 𝑘𝜎) ≥ 1 − Sample Random Variables (Statistics)
𝑘2
Standardized Variable Sample data 𝑥1 , 𝑥2 , … , 𝑥𝑛
𝑋−𝜇 Each independent measures of RV 𝑋𝑖
𝑍= 1
𝜎
𝑥−𝜇 𝑥̅ = ∑𝑛𝑖=1 𝑥𝑖 (empirical value of mean)
𝑥−𝜇 𝑛
𝑃 (𝑍 ≤ ) = ∫−∞ 𝑛(𝑠; 0,1) 𝑑𝑠
𝜎
1
𝜎 𝑋̅ = ∑𝑛𝑖=1 𝑋𝑖 (sample mean RV)
𝐴−𝜇 𝐵−𝜇 𝐵−𝜇 𝐴−𝜇 𝑛
𝑃( ≤𝑍≤ ) = Φ( ) − Φ( ) Data in increasing order 𝑥(1) , … , 𝑥(𝑛)
𝜎 𝜎 𝜎 𝜎
𝑥−𝜇 𝑥 𝑛 +𝑥 𝑛
𝑛( 𝜎 ;0,1) ( ) ( +1)
𝑛(𝑥; 𝜇, 𝜎) = 𝑀𝑒𝑑𝑖𝑎𝑛 = 2 2
, n is even
𝜎 2
Normal Approx. of Binomial Distribution 𝑀𝑒𝑑𝑖𝑎𝑛 = 𝑥(𝑛+1) , n is odd
𝑥+0.5−𝑛𝑝 2
𝑃(𝑋 ≤ 𝑥) ≈ 𝑃 (𝑍 ≤ ) 1
√𝑛𝑝(1−𝑝) 𝑠2 = ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )2 (sample variance)
𝑛−1
𝑛𝑝 ≥ 5, 𝑛(1 − 𝑝) ≥ 5
𝑠 = √𝑠 2 (sample standard deviation)
Gamma Function
∞
𝑅𝑎𝑛𝑔𝑒 = max(𝑥𝑖 ) − min (𝑥𝑖 )
Γ(𝛼) = ∫0 𝑥 𝛼−1 𝑒 −𝑥 𝑑𝑥 , 𝛼 > 0 𝐸[𝑋̅] = 𝜇 and 𝐸[𝑆 2 ] = 𝜎 2 (unbiased)
1 1
Γ ( ) = √𝜋, Γ(𝑛) = (𝑛 − 1)!, 𝑛 ∈ ℕ
2
𝑆2 = [𝑛 ∑𝑛𝑖=1 𝑋𝑖2 − (∑𝑛𝑖=1 𝑋𝑖 )2 ]
𝑛(𝑛−1)
Poisson and Exponential Distributions Box-and-Whisker and Quantile Plots
𝑑 −𝜆𝑥 1 𝑖
𝑃(𝑋 ≤ 𝑥) = 𝜆𝑒 ,𝜆 = 𝑄𝑖 = (𝑛 + 1) ⋅ , 𝑖 = 1, 2, 3 (quartiles)
𝑑𝑥 𝛽 4
Transformations of Random Variables 𝐼𝑄𝑅 = 𝑄3 − 𝑄1 , 𝑄2 is the median
𝑌 = 𝑢(𝑋) ∴ 𝑢−1 (𝑌) exists if bijective Lower Whisker (minimum): 𝑄1 − 1.5(𝐼𝑄𝑅)
Given distribution 𝑓(𝑥) Upper Whisker (maximum): 𝑄3 + 1.5(𝐼𝑄𝑅)
𝑔(𝑦) = 𝑓(𝑢−1 (𝑦)) (Discrete) 𝑖−8
3

𝑑𝑢−1 (𝑦) Quantile Plot: ( 1 , 𝑥𝑖 ) = (𝑓, 𝑥𝑖 )

𝑔(𝑦) = 𝑓(𝑢−1 (𝑦)) ⋅ | | (Continuous) 𝑛+
4
𝑑𝑦
𝑞𝜇,𝜎 (𝑓) = 𝜇 + 𝜎{4.91[𝑓 0.14 − (1 − 𝑓)0.14 ]} (Normal Q-Q) =
𝑌1 = 𝑢1 (𝑋1 , 𝑋2 ), 𝑌2 = 𝑢2 (𝑋1 , 𝑋2 )
(𝑞0,1 (𝑓𝑖 ), 𝑥𝑖 )
∴ 𝑋1 = 𝑢1−1 (𝑌1 , 𝑌2 ), 𝑋2 = 𝑢2−1 (𝑌1 , 𝑌2 )
𝑞0,1 = 4.91[𝑓 0.14 − (1 − 𝑓)0.14 ]
𝑔(𝑦1 , 𝑦2 ) = 𝑓(𝑢1−1 (𝑦1 , 𝑦2 ), 𝑢2−1 (𝑦1 , 𝑦2 )|𝐽| (Continuous)
𝜕𝑥1 𝜕𝑥1 For CDF 𝐹(𝑥), 𝑞(𝑓) = 𝐹 −1
𝜕𝑦 𝜕𝑦2 CLT and T-Distribution (Mean), Chi-Squared (Variance)
|𝐽| = |𝜕𝑥1 𝜕𝑥2
|
2 𝑋̅𝑛−𝜇
𝜕𝑦1 𝜕𝑦2 𝑍𝑛 = as 𝑛 → ∞, 𝑍𝑛 → 𝑛(𝑧; 0,1), 𝑛 ≥ 30
𝜎/√𝑛
𝑔(𝑦1 , 𝑦2 ) = 𝑓(𝑢1−1 (𝑦1 , 𝑦2 ), 𝑢2−1 (𝑦1 , 𝑦2 )) (Discrete) 𝑋̅−𝜇
𝑇= , 𝑛 < 30, 𝜈 = 𝑛 − 1
𝑆/√𝑛
𝑔(𝑦) = ∑𝑘𝑖=1 𝑓[𝑤𝑖 (𝑦)]|𝐽𝑖 | (non-bijective)
(𝑛−1)𝑆 2
𝜒2 =
𝜎2
Winter 2024 Final Exam
ECE286H1 S
Duration: 150 Minutes

Point Estimates 𝑝̂(1−𝑝̂) 𝑝̂(1−𝑝̂)

𝑃 (𝑝̂ − 𝑧𝛼 √ ≤ 𝑝 ≤ 𝑝̂ + 𝑧𝛼 √ )=1−𝛼
𝜃 = 𝜇, 𝜃̂ = 𝑥̅ , Θ
̂ = 𝑋̅ 2 𝑛 2 𝑛

̂
𝐸[Θ] = 𝜃 unbiased estimator 2 ̂(1−𝑝̂)
𝑧𝛼 𝑝
2
Confidence Intervals 𝑛=
𝛿2
2
𝑃(Θ𝐿 ≤ 𝜃 ≤ Θ𝑈 ) = 1 − 𝛼 𝑧𝛼
𝜎 𝜎 𝑛≥ 2
, max(𝑝̂ (1 − 𝑝̂ ) = 0.25
𝑃 (𝑋̅ − 𝑧𝛼 ≤ 𝜇 ≤ 𝑋̅ + 𝑧𝛼 )=1−𝛼 4𝛿 2
2 √𝑛 2 √𝑛 CI For Variance
𝑠 𝑠
𝑃 (𝑋̅ − 𝑡𝛼 ≤ 𝜇 ≤ 𝑋̅ + 𝑡𝛼 )=1−𝛼 (𝑛−1)𝑆 2 (𝑛−1)𝑆 2
2 √𝑛 2 √𝑛 𝑃( ≤ 𝜎2 ≤ ) = 1 − 𝛼, 𝜈 = 𝑛 − 1
𝜎 𝜎 χ2 𝜒2 𝛼
𝑒= (standard), 𝑒 = 𝑧𝛼 (margin) 𝛼
2 1− 2
√𝑛 2 √𝑛
𝜎 Maximum Likelihood Estimation and Log-Likelihood
𝑃 (𝜇 ≤ 𝑋̅ + 𝑧𝛼 ) = 1 − 𝛼 (upper bound)
√𝑛 Samples 𝑥1 , … , 𝑥𝑛 , Joint PDF 𝑓(𝑥1 , … , 𝑥𝑛 ; 𝜃)
𝜎
𝑃 (𝜇 ≥ 𝑋̅ − 𝑧𝛼 ) = 1 − 𝛼 (lower bound) 𝐿(𝑥1 , … , 𝑥𝑛 ; 𝜃) = ∏𝑛𝑖=1 𝑔(𝑥𝑖 ; 𝜃)
√𝑛
𝑑𝐿 𝑑 2𝐿
Prediction Intervals 𝜃̂ = arg max 𝐿(𝑥1 , … , 𝑥𝑛 ; 𝜃) = 𝜃 𝑠. 𝑡. = 0 and <0
𝑑𝜃 𝑑𝜃 2
For next observation 𝑥0 log 𝐿 = log(∏𝑛𝑖=1 𝑔(𝑥𝑖 ; 𝜃)) = ∑𝑛𝑖=1 log(𝑔(𝑥𝑖 ; 𝜃))
1 1 𝑑(log 𝐿)
𝑃 (𝑥̅ − 𝑧𝛼 𝜎√1 + ≤ 𝑥0 ≤ 𝑥̅ + 𝑧𝛼 𝜎√1 + ) = 1 − 𝛼
𝑛 𝑛
𝜃̂ = arg max log 𝐿(𝑥1 , … , 𝑥𝑛 ; 𝜃) = 𝜃 𝑠. 𝑡. =0
2 2 𝑑𝜃
𝜕 1 𝑛
Tolerance Limits 𝜇= ln(𝐿(𝑥1 , … , 𝑥𝑛 ; 𝜇, 𝜎)) = 𝑥̅ = ∑ 𝑥
𝜕𝜇 𝑛 𝑖=1 𝑖
𝑃(𝑥̅ ± 𝑘𝑠) = 1 − 𝛾 (that 1 − 𝛼 of samples in range) 𝜕 𝑛−1 2 1
𝜎2 = ln(𝐿(𝑥1 , … , 𝑥𝑛 ; 𝜇, 𝜎)) = 𝑠 = ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )2
CI for Difference of Means with Known 𝝈 𝜕𝜎 2 𝑛 𝑛
Hypothesis Testing (One and Two-Tailed)
𝜎12 𝜎22
𝑃 ((𝑥̅1 − 𝑥̅2 ) − 𝑧𝛼 √ + ≤ 𝜇1 − 𝜇2 ≤ (𝑥̅1 − 𝑥̅2 ) + Null: 𝐻0 : 𝜃 = 𝜃0
2 𝑛1 𝑛2
Alternate: 𝐻1 : 𝜃 ≠ 𝜃0 (two-sided), 𝐻1 : 𝜃 > 𝜃0 (upper bound),
𝜎12 𝜎22 𝐻1 : 𝜃 < 𝜃0 (lower bound)
𝑧𝛼 √ + ) = 1−𝛼
2 𝑛1 𝑛2 Type I (False Positive) and Type II (False Negative) Errors
Unknown and Equal Population Variance 𝛼 = 𝑃(𝑡𝑦𝑝𝑒 𝐼) = 1 − 𝑃(𝐶𝑟𝑖𝑡𝑖𝑐𝑎𝑙 𝑅𝑒𝑔𝑖𝑜𝑛) with 𝐻0 parameter
(𝑛1 −1)𝑆12 +(𝑛2 −1)𝑆22 (level of significance)
𝑆𝑝2 = , 𝑛1 , 𝑛2 < 30, 𝜎1 = 𝜎2 𝛽 = 𝑃(𝑡𝑦𝑝𝑒 𝐼𝐼) = 𝑃(𝐶𝑟𝑖𝑡𝑖𝑐𝑎𝑙 𝑅𝑒𝑔𝑖𝑜𝑛) with 𝐻1 parameter
𝑛1 +𝑛2 −2

1 1 𝑃𝑜𝑤𝑒𝑟 = 1 − 𝛽
𝑃 ((𝑥̅1 − 𝑥̅2 ) − 𝑡𝛼 𝑆𝑝 √ + ≤ 𝜇1 − 𝜇2 ≤ (𝑥̅1 − 𝑥̅2 ) +
2 𝑛1 𝑛2 P-Value and Critical Region (Symmetric)
1 1 𝑃 ≤ 𝛼, reject 𝐻0 , 𝑃 ≥ 𝛼, fail to reject 𝐻0
𝑡𝛼 𝑆𝑝 √ + ) = 1 − 𝛼, 𝜈 = 𝑛1 + 𝑛2 − 2
2 𝑛1 𝑛2 𝑃 = 2𝑃(𝑀 > |𝑚|) (two-sided, 𝑀 = 𝑍, 𝑇, 𝑚 = 𝑧, 𝑡)
Unknown and Unequal Population Variance 𝑃 = 𝑃(𝑀 > |𝑚|) (one-sided, 𝑀 = 𝑍, 𝑇, 𝑚 = 𝑧, 𝑡)
𝑆12 𝑆22 𝑃 (−𝑚𝛼 ≤ 𝑚 ≤ 𝑚𝛼 ) = 1 − 𝛼, reject 𝐻0 if |𝑚| > 𝑚𝛼 ,
𝑃 ((𝑥̅1 − 𝑥̅2 ) − 𝑡𝛼 √ + ≤ 𝜇1 − 𝜇2 ≤ (𝑥̅1 − 𝑥̅2 ) + 2 2 2
𝑛1 𝑛2
2 otherwise fail to reject 𝐻0 (two-sided)
𝑆2 𝑆 2 2 𝑃(𝑚 ≤ 𝑚𝛼 ) = 1 − 𝛼 or 𝑃(𝑚 ≥ 𝑚𝛼 ) = 1 − 𝛼, reject 𝐻0 if
( 1+ 2)
𝑆12 𝑆22 𝑛1 𝑛2 |𝑚| > 𝑚𝛼 , otherwise fail to reject 𝐻0
𝑡𝛼 √ + ) = 1 − 𝛼, 𝜈 = 2 2
2 𝑛1 𝑛2 𝑆2 𝑆2
(𝑛1 ) (𝑛2 )
P-Value and Critical Region (Asymmetric)
1 2
⌊ ⌋ +
𝑛1 −1 𝑛2 −1 𝑃 ≤ 𝛼, reject 𝐻0 , 𝑃 ≥ 𝛼, fail to reject 𝐻0
CI for Paired Observations 𝑃 = 𝑃(𝑅 > 𝑟) + 𝑃(𝑅 < 𝑟) (two-sided, 𝑅 = 𝜒 2 , 𝐹, 𝑟 = 𝜒 2 , 𝑓)
𝐷𝑖 = 𝑋1,𝑖 − 𝑋2,𝑖 𝑃 = 𝑃(𝑅 > 𝑟) or 𝑃(𝑅 < 𝑟) (one-sided, 𝑅 = 𝜒 2 , 𝐹, 𝑟 = 𝜒 2 , 𝑓)
𝜇𝐷 = 𝜇1 − 𝜇2 , 𝑑̅ = 𝑥̅1 − 𝑥̅2 , 𝜈 = 𝑛 − 1 𝑃 (𝑟𝛼 ≤ 𝑟 ≤ 𝑟1−𝛼 ) = 1 − 𝛼, reject 𝐻0 if 𝑟 < 𝑟1−𝛼 or
𝑣𝑎𝑟(𝐷𝑖 ) = 𝜎𝑋21,𝑖 + 𝜎𝑋22,𝑖 − 2𝑐𝑜𝑣(𝑋1,𝑖 , 𝑋2,𝑖 ) 2 2 2

𝑠𝑑 𝑠𝑑
𝑟 > 𝑟 , otherwise fail to reject 𝐻0 (two-sided)
𝛼
𝑃 (𝑑̅ − 𝑡𝛼 ≤ 𝜇𝐷 ≤ 𝑑̅ + 𝑡𝛼 )=1−𝛼 2
2 √𝑛 2 √𝑛 𝑃(𝑟 ≤ 𝑟1−𝛼 ) = 1 − 𝛼 or 𝑃(𝑟 ≥ 𝑟𝛼 ) = 1 − 𝛼, reject 𝐻0 if
CI for Estimating a Proportion (Single Sample) 𝑟 > 𝑟𝛼 or 𝑟 < 𝑟1−𝛼 , otherwise fail to reject 𝐻0
𝑋 𝑥
𝑃̂ = , 𝑝̂ = , 𝑝 unknown (Binomial distribution) Hypothesis Testing for the Mean (Known 𝝈𝟐 )
𝑛 𝑛
𝐻0 : 𝜇 = 𝜇0 , 𝐻1 : 𝜇 ≠ 𝜇0 (two-sided), 𝐻1 : 𝜇 > 𝜇0 or 𝐻1 : 𝜇 < 𝜇0
(one-sided)
Winter 2024 Final Exam
ECE286H1 S
Duration: 150 Minutes
𝑋̅ −𝜇0 𝜕(𝑆𝑆𝐸) 𝜕(𝑆𝑆𝐸)
𝑧= , 𝑛 ≥ 30 𝑆𝑆𝐸 = ∑𝑛𝑖=1 𝑒𝑖2 , = 0, =0
𝜎/√𝑛 𝜕𝑏0 𝜕𝑏1
1
𝑃 = 2𝑃(𝑍 > |𝑧|) = 2(1 − Φ(|𝑧|)) = 2Φ(−|𝑧|) (two-sided) 𝑏0 = 𝑦̅ − 𝑏1 𝑥̅ = (∑𝑛𝑖=1 𝑦𝑖 − 𝑏1 ∑𝑛𝑖=1 𝑥𝑖 )
𝑛
𝑃 = 𝑃(𝑍 > |𝑧|) = 1 − Φ(|𝑧|) (one-sided) ∑𝑛
𝑖=1(𝑥𝑖 −𝑥̅ )(𝑦𝑖 −𝑦
̅) 𝑛 ∑𝑛 𝑛 𝑛
𝑖=1 𝑥𝑖 𝑦𝑖 −(∑𝑖=1 𝑥𝑖 )(∑𝑖=1 𝑦𝑖 )
𝑏1 = ∑𝑛
= 2
Hypothesis Testing for the Mean (Unknown 𝝈𝟐 ) 𝑖=1(𝑥𝑖 −𝑥̅ )
2
𝑛 ∑𝑛 2 𝑛
𝑖=1 𝑥𝑖 −(∑𝑖=1 𝑥𝑖 )

𝑡=
𝑋̅ −𝜇0
, 𝜈 = 𝑛 − 1, 𝑛 ≤ 30 Sums of Errors
𝑠/√𝑛 1
𝑆𝑥𝑥 = ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )2 = ∑𝑛𝑖=1 𝑥𝑖2 − (∑𝑛𝑖=1 𝑥𝑖 )2
Hypothesis Testing for Two Means (Known 𝝈𝟐 ) 𝑛

𝐻0 : 𝜇1 − 𝜇2 = 𝑑0 𝐻1 : 𝜇1 − 𝜇2 ≠ 𝑑0 (two-sided) 𝑆𝑥𝑥 = ∑𝑛𝑖=1 𝑥𝑖2 − 𝑛(𝑥̅ )2

1
𝐻1 : 𝜇1 − 𝜇2 > 𝑑0 or 𝐻1 : 𝜇1 − 𝜇2 < 𝑑0 (one-sided) 𝑆𝑦𝑦 = ∑𝑛𝑖=1(𝑦𝑖 − 𝑦̅)2 = ∑𝑛𝑖=1 𝑦𝑖2 − (∑𝑛𝑖=1 𝑦𝑖 )2
𝑛
(𝑥̅ 1 −𝑥̅2 )−𝑑0
𝑧= (CLT) 𝑆𝑦𝑦 = ∑𝑛𝑖=1 𝑦𝑖2 − 𝑛(𝑦̅)2
𝜎2 𝜎2
√ 1+ 2
𝑛1 𝑛2 𝑆𝑥𝑦 = ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅)
1
Hypothesis Testing for Two Means (Unknown 𝝈𝟐𝟏 = 𝝈𝟐𝟐 ) 𝑆𝑥𝑦 = ∑𝑛𝑖=1 𝑥𝑖 𝑦𝑖 − (∑𝑛𝑖=1 𝑥𝑖 )(∑𝑛𝑖=1 𝑦𝑖 ) = ∑𝑛𝑖=1 𝑥𝑖 𝑦𝑖 − 𝑛𝑥̅ 𝑦̅
𝑛
(𝑥̅1 −𝑥̅2 )−𝑑0 (𝑛1 −1)𝑆12 +(𝑛2 −1)𝑆22 𝑆𝑥𝑦 2
𝑆𝑥𝑦
𝑡= 1 1
, 𝜈 = 𝑛1 + 𝑛2 − 2, 𝑆𝑝2 = 𝑆𝑆𝐸 = 𝑆𝑦𝑦 − 𝑏1 𝑆𝑥𝑦 = 𝑆𝑦𝑦 − ( ) 𝑆𝑥𝑦 = 𝑆𝑦𝑦 −
𝑠𝑝 √ + 𝑛1 +𝑛2 −2 𝑆𝑥𝑥 𝑆𝑥𝑥
𝑛1 𝑛2
2 2] 𝑆𝑆𝐸
𝑠 = 𝐸[𝜎 =
Hypothesis Testing for Two Means (Unknown 𝝈𝟐𝟏 ≠ 𝝈𝟐𝟐 ) 𝑛−2
2 CI for Regression Parameters
𝑆2 𝑆2
( 1+ 2) 𝑏1 −𝛽1 𝑏0 −𝛽0
(𝑥̅1 −𝑥̅2 )−𝑑0 𝑛1 𝑛2 𝑡= ,𝑡=
𝑡= ,𝜈= 2 2 𝑠/√𝑆𝑥𝑥 ∑ 𝑥 𝑛 2
2 2 𝑆2 𝑆2 𝑠⋅√ 𝑖=1 𝑖
𝑠 𝑠
√ 1+ 2 (𝑛1 ) (𝑛2 ) 𝑛𝑆𝑥𝑥
𝑛1 𝑛2 1 2
⌊ ⌋ +
𝑛1 −1 𝑛2 −1
𝑠 𝑠
Hypothesis Testing for Paired Observations 𝑃 (𝑏1 − 𝑡𝛼 ≤ 𝛽1 ≤ 𝑏1 + 𝑡𝛼 ) = 1 − 𝛼, 𝜈 = 𝑛 − 2
2 √𝑆𝑥𝑥 2 √𝑆𝑥𝑥
𝐻0 : 𝜇𝐷 = 𝑑0 𝐻1 : 𝜇𝐷 ≠ 𝑑0 (two-sided) 𝑠 𝑠
𝑃 (𝑏0 − 𝑡𝛼 √∑𝑛𝑖=1 𝑥𝑖2 ≤ 𝛽0 ≤ 𝑏0 + 𝑡𝛼 √∑𝑛𝑖=1 𝑥𝑖2 ) =
𝐻1 : 𝜇𝐷 > 𝑑0 or 𝐻1 : 𝜇𝐷 < 𝑑0 (one-sided) 2 √𝑛𝑆𝑥𝑥 2 √𝑆𝑥𝑥
𝑑̅ −𝑑0 1 − 𝛼, 𝜈 = 𝑛 − 2
𝑡= ,𝜈 =𝑛−1
𝑠𝑑 /√𝑛 Hypothesis Testing for Regression Parameters
Hypothesis Testing for One Variance 𝐻0 : 𝛽1 = 𝛽1 0 , 𝐻1 : 𝛽1 ≠ 𝛽1 0 (two-sided)
𝐻0 : 𝜎 2 = 𝜎02 , 𝐻1 : 𝜎 2 ≠ 𝜎02 (two-sided) 𝐻1 : 𝛽1 > 𝛽1 0 or 𝐻1 : 𝛽1 < 𝛽1 0 (one-sided)
𝐻1 : 𝜎 2 > 𝜎02 or 𝐻1 : 𝜎 2 < 𝜎02 (one-sided) 𝐻0 : 𝛽0 = 𝛽0 0 , 𝐻1 : 𝛽0 ≠ 𝛽0 0 (two-sided)
(𝑛−1)𝑠 2
𝜒2 = ,𝜈 =𝑛−1 𝐻1 : 𝛽0 > 𝛽0 0 or 𝐻1 : 𝛽0 < 𝛽0 0 (one-sided)
𝜎02
𝑏1 −𝛽1 𝑏0 −𝛽0
Hypothesis Testing for Two Variances 𝑡= ,𝑡=
𝑠/√𝑆𝑥𝑥 𝑛 2
𝐻0 : 𝜎12 = 𝜎22 , 𝐻1 : 𝜎12 ≠ 𝜎22 (two-sided) ∑ 𝑥
𝑠⋅√ 𝑖=1 𝑖
𝑛𝑆𝑥𝑥
𝐻1 : 𝜎12 > 𝜎22 or 𝐻1 : 𝜎12 < 𝜎22 (one-sided)
𝑠12
Coefficient of Determination 𝑹𝟐
𝑓= 2 , 𝜈1 = 𝑛1 − 1, 𝜈2 = 𝑛2 − 1 𝑆𝑆𝐸 (𝑦 −𝑦̂𝑖 )2
𝑠2 𝑅2 = 1 − = 1 − ∑𝑛𝑖=1 (𝑦𝑖 ̅) 2 , 0 ≤ 𝑅2 ≤ 1
1 𝑆𝑆𝑇 𝑖 −𝑦
𝑓1−𝛼 (𝜈1 , 𝜈2 ) =
𝑓𝛼 (𝜈2 ,𝜈1 )
Goodness of Fit
𝑒𝑖 = 𝑛𝑃(𝑖), n trials, k outcomes
(𝑂𝑖 −𝑒𝑖 )2
𝜒 2 = ∑𝑘𝑖=1 ,𝜈 =𝑘−1
𝑒𝑖
Small 𝜒 2 = good fit
Reject if 𝜒 2 > 𝜒𝛼2
Linear Regression
Data {(𝑥𝑖 , 𝑦𝑖 ); 𝑖 = 1,2, … , 𝑛}
Real: 𝑦𝑖 = 𝛽0 + 𝛽1 𝑥𝑖 + 𝜖𝑖 , 𝐸[𝜖] = 0, 𝑉𝑎𝑟(𝜖) = 𝜎 2
Estimator: 𝑦̂𝑖 = 𝑏0 + 𝑏1 𝑥𝑖
Residual: 𝑒𝑖 = 𝑦𝑖 − 𝑦̂𝑖 , 𝑖 = 1, … , 𝑛
𝑦𝑖 = 𝑏0 + 𝑏1 𝑥𝑖 + 𝑒𝑖