Mid-term Exam Formula Sheet
BASIC CONCEPTS
Summation: ∑𝑁 𝑁 𝑁
𝑖=1(𝑎 + 𝑏𝑥𝑖 + 𝑐𝑦𝑖 ) = 𝑁𝑎 + 𝑏 ∑𝑖=1 𝑥𝑖 + 𝑐 ∑𝑖=1 𝑦𝑖 , where a, b, and c are constants.
∑𝑁 2 𝑁 2
𝑖=1(𝑥𝑖 − 𝑥̄ ) = ∑𝑖=1 𝑥𝑖 − 𝑁𝑥̄
2
∑𝑁 𝑁 𝑁 𝑁
𝑖=1(𝑥𝑖 − 𝑥̄ )(𝑦𝑖 − 𝑦̄ ) = ∑𝑖=1 𝑥𝑖 (𝑦𝑖 − 𝑦̄ ) = ∑𝑖=1(𝑥𝑖 − 𝑥̄ )𝑦𝑖 = ∑𝑖=1 𝑥𝑖 𝑦𝑖 − 𝑁𝑥̄ 𝑦̄
𝑏
Probability distribution: pdf - ∫𝑎 𝑓𝑌 (𝑦)𝑑𝑦 = 𝑃(𝑎 ≤ 𝑌 ≤ 𝑏); cdf - F(y) = P(Y ≤ y );
joint distribution - 𝑓𝑋,𝑌 (𝑥, 𝑦); conditional distribution - 𝑓𝑌|𝑋 (𝑦|𝑥) = 𝑓𝑋,𝑌 (𝑥, 𝑦)/𝑓𝑋 (𝑥)
∞
Expected value: 𝐸(𝑌) = ∫−∞ 𝑦𝑓𝑌 (𝑦)𝑑𝑦 = 𝜇𝑌 ; 𝐸(∑𝑁 𝑁
𝑖=1 𝑎𝑖 𝑌𝑖 ) = ∑𝑖=1 𝐸(𝑎𝑖 𝑌𝑖 )
Variance: 𝑉𝑎𝑟(𝑌) = 𝐸[(𝑌 − 𝜇𝑌 )]2 = 𝜎𝑌2 ; Var(∑𝑁 𝑁 2
𝑖=1 𝑎𝑖 𝑌𝑖 ) = ∑𝑖=1 𝑎𝑖 Var(𝑌𝑖 ) + 2 ∑𝑖<𝑗 ∑ 𝑎𝑖 𝑎𝑗 Cov(𝑌𝑖 , 𝑌)
Standard deviation: 𝜎𝑌 ; Skewness: 𝐸[(𝑌 − 𝜇𝑌 )]3 /𝜎𝑌3 ; Kurtosis: 𝐸[(𝑌 − 𝜇𝑌 )]4 /𝜎𝑌4
Covariance: 𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸[(𝑋 − 𝜇𝑋 )(𝑌 − 𝜇𝑌 )] = 𝜎𝑋𝑌
Cov(a + bX, c + dY) = bdCov(X, Y), where a, b, c and d are constants
Correlation coefficient: 𝜌𝑋𝑌 = 𝐶𝑜𝑣(𝑋, 𝑌)/𝜎𝑋 𝜎𝑌
Corr(a + bX, c + dY) = Corr(X, Y), where a, b, c and d are constants
𝑌−𝜇𝑌
Normal distribution: 𝑌~𝑁𝑜𝑟𝑚𝑎𝑙(𝜇𝑌 , 𝜎𝑌2 ); Standard normal distribution: 𝑍 = 𝜎𝑌
~𝑁𝑜𝑟𝑚𝑎𝑙(0,1)
2
𝑍 𝐾𝑁 /𝑁1
𝑖=1 𝑍𝑖 ~𝑁 ; t-distribution: 𝑇 =
Chi-square distribution: 𝐾 = ∑𝑁 2 2
~𝑡𝑁 ; F-distribution: F = 𝐾2 1 /𝑁 = 𝐹𝑁1 ,𝑁2
√/𝑁 𝑁2 2
𝑌𝑖
Sample mean: 𝑌̅ = ∑𝑁 ̂𝑌2 = ∑𝑁
𝑖=1 𝑁 ; Sample variance: 𝜎
̅ 2
𝑖=1(𝑌𝑖 − 𝑌) /(𝑁 − 1)
Sample covariance: 𝐶𝑜̂𝑣(𝑋, 𝑌) = [1/(𝑁 − 1)] ∑𝑁𝑖=1(𝑋𝑖 − 𝑋̄)(𝑌𝑖 − 𝑌̄)
Unbiasedness: 𝐸(𝜃̂) = 𝜃 ; Consistency: 𝑙𝑖𝑚 𝑃(|𝜃̂𝑁 − 𝜃| > ∈) = 0, ∈ > 0
𝑁→∞
SIMPLE LINEAR REGRESSION MODEL
Econometric model/Population regression function (PRF): Yi = E(Y|Xi) + ui = 0 + 1Xi + ui, 𝑢𝑖 ~𝑖𝑖𝑑 𝑁𝑜𝑟𝑚𝑎𝑙(0, 𝜎 2 )
̂ 𝑖 ) + 𝑢̂𝑖 = 𝑌̂𝑖 + 𝑢̂𝑖 = 𝛽̂0 + 𝛽̂1 𝑋𝑖 + 𝑢̂𝑖
Estimated (fitted) model/Sample regression function (SRF): 𝑌𝑖 = 𝐸(𝑌|𝑋
𝑁
OLS estimator: 𝛽̂0 = 𝑌 ̂ = ∑𝑖=1(𝑋𝑖−𝑋̅)(𝑌𝑖−𝑌̅)
̅ − 𝛽̂1 𝑋̅; 𝛽
1 𝑁 2
̅)
∑𝑖=1(𝑋𝑖 −𝑋
𝑁 2 ∑𝑁
𝜎 ∑ 𝑋𝑖 2 𝜎2 ̂𝑖2
𝑖=1 𝑢
𝑉𝑎𝑟(𝛽̂0 ) = 𝑁 ∑𝑁 𝑖=1
(𝑋 −𝑋̅ )2
; V𝑎𝑟(𝛽̂1 ) = ∑𝑁 ̅ 2
; 𝜎̂ 2 =
𝑖=1 𝑖 𝑖=1(𝑋𝑖 −𝑋 ) 𝑁−2
2
Goodness-of-fit: ∑𝑁 ̅ 2 𝑁 ̂ ̅ 𝑁
̂ 𝑖2 ;
𝑖=1( 𝑌𝑖 − 𝑌 ) = ∑𝑖=1(𝑌𝑖 − 𝑌 ) + ∑𝑖=1 𝑢
(SST = SSE + SSR)
𝑆𝑆𝐸 𝑆𝑆𝑅
Coefficient of determination: 𝑅 2 = = 1 − 𝑆𝑆𝑇
𝑆𝑆𝑇
̂𝑘 −𝛽 0
𝛽 𝑘
t-test: 𝑡 = ̂𝑘 ) ~𝑡(𝑁−2) , k = 0, 1
𝑠𝑒(𝛽
Confidence interval: 𝑃[𝛽̂𝑘 − 𝑡𝑑𝑓,𝛼 se(𝛽̂𝑘 ) ≤ 𝛽𝑘 ≤ 𝛽̂𝑘 + 𝑡𝑑𝑓,𝛼 se(𝛽̂𝑘 )] = 1 − 𝛼
𝑆2 (𝐾−3)2
Jarque-Bera (JB) test: 𝐽𝐵 = 𝑁 [ + ] ~𝑎𝑠𝑦 22
6 24
