Lecture 4: Proofs for Expectation, Variance, and Covariance Formula

Transcription

1 Lecture 4: Proofs for Expectation, Variance, and Covariance Formula by Hiro Kasahara Vancouver School of Economics University of British Columbia Hiro Kasahara (UBC) Econ / 28

2 Discrete Random Variables: X and Y Let X and Y be two discrete random variables. X takes n possible values: {x 1,..., x n } Y takes m possible values: {y 1,..., y m }. Hiro Kasahara (UBC) Econ / 28

3 Notations: The joint probability mass function is given by p X,Y ij = P (X = x i, Y = y j ), i = 1,... n; j = 1,..., m. The marginal probability mass function of X is p X i = P (X = x i ) = m j=1 p X,Y ij, i = 1,... n, and the marginal probability mass function of Y is p Y j = P (Y = y j ) = n p X,Y ij, j = 1,... m. Hiro Kasahara (UBC) Econ / 28

4 Notations: Table: Example of Joint Distribution of X and Y Y = y 1 Y = y 2 Y = y 3 Marg. prob. of X X = x 1 p X,Y 11 p X,Y 12 p X,Y 13 p1 X X = x 2 p X,Y 21 p X,Y 22 p X,Y 23 Marg. prob. of Y p1 Y p2 Y p3 Y p2 X 1.00 P (X = x 1 ) = 3 j=1 p X,Y 1j = p X,Y 11 + p X,Y 12 + p X,Y 13 = p X 1. Hiro Kasahara (UBC) Econ / 28

5 Notations: Table: Example of Joint Distribution of X and Y Y = 30 Y = 60 Y = 100 Marg. Prob. of X X = X = Marg. Prob. of Y P(X = 0) = = Hiro Kasahara (UBC) Econ / 28

6 Proposition If a and b are constants, then E(a + bx) = a + be(x). Proof: E(a + bx) n def = (a + bx i )pi X = (a + bx 1 )p X (a + bx n )p X n = (ap X ap X n ) + (bx 1 p X bx n p X n ) = a (p1 X pn X ) + b (x 1 p1 X x n pn X ) n n = a pi X + b x i pi X = a + be(x). Hiro Kasahara (UBC) Econ / 28

7 Clicker Question 4-1 For any constant a, b, and for any function g(x), which of the following is true? A). E[a + b g(x)] = a + b g(e[x]). B). E[a + b g(x)] = a + b E[g(X)]. C). E[a + b g(x)] = a + b g(e[x]) = a + b E[g(X)]. Hiro Kasahara (UBC) Econ / 28

8 Question 1 in Worksheet Prove that, for any any constant a, b, and for any function g(x), E[a + b g(x)] = a + b E[g(X)]. Hiro Kasahara (UBC) Econ / 28

9 Proposition Cov (a 1 + b 1 X, a 2 + b 2 Y ) = b 1 b 2 Cov (X, Y ), where a 1, a 2, b 1, and b 2 are some constants. Proof: Cov(X, Y ) def = E[(a 1 + b 1 X E(a 1 + b 1 X))(a 2 + b 2 Y E(a 2 + b 2 Y ))] = E[(a 1 a 1 + b 1 X b 1 E(X))(a 2 a 2 + b 2 Y b 2 E(Y )] = E[b 1 b 2 (X E(X))(Y E(Y ))] = n m j=1 = b 1 b 2 n b 1 b 2 (x i E(X))(y j E(Y )) p X,Y ij m j=1 = b 1 b 2 Cov(X, Y ). [x i E(X)][y j E(Y )] p X,Y ij Hiro Kasahara (UBC) Econ / 28

10 Clicker Question 4-2 For any constant a and b, which of the following is true? A). Var[a + bx] = a + bvar[x]. B). Var[a + bx] = bvar[x]. C). Var[a + bx] = b 2 Var[X]. Hiro Kasahara (UBC) Econ / 28

11 Question 2 in Worksheet Prove that, for any any constant a, b, Var[a + bx] = b 2 Var[X]. Hiro Kasahara (UBC) Econ / 28

12 Proposition If X and Y are independent, then Cov (X, Y ) = 0. Proof: Cov(X, Y ) def = E[(X E(X))(Y E(Y ))] n m = [x i E(X)][y j E(Y )] P(X = x i, Y = y j ) = = n n j=1 m [x i E(X)][y j E(Y )]pi X pj Y (by independence) j=1 j=1 m {[x i E(X)]pi X }{[y j E(Y )]pj Y } Hiro Kasahara (UBC) Econ / 28

13 Proof continued = = n m {[x i E(X)]pi X }{[y j E(Y )]pj Y } j=1 n [x i E(X)]pi X { n = x i pi X = { E(X) E(X) m [y j E(Y )]pj Y j=1 } n m E(X)pi X y j pj Y n p X i } j=1 m E(Y ) E(Y ) (by definition of E(X) and E(Y )) j=1 p Y j m E(Y )pj Y j=1 Hiro Kasahara (UBC) Econ / 28

14 Proof continued = { E(X) E(X) n p X i } m E(Y ) E(Y ) = {E(X) E(X) 1} {E(Y ) E(Y ) 1} = 0 0 = 0. j=1 p Y j Hiro Kasahara (UBC) Econ / 28

15 Clicker Question 4-3 Suppose that X and Y are independent. For any function g(x) and h(y), which of the following is true? A). Cov (g(x), h(y )) = 0 always holds. B). Cov (g(x), h(y )) = 0 does not hold in some cases. Hiro Kasahara (UBC) Econ / 28

16 Question 4 in Worksheet Prove that, when X and Y are independent, for any function g(x) and h(y). Cov (g(x), h(y )) = 0 Hiro Kasahara (UBC) Econ / 28

17 Proposition Var (X + Y ) = Var (X) + Var (Y ) + 2Cov (X, Y ). Proof: Let X def = X E(X) and Ỹ def = Y E(Y ). Var(X + Y ) def = E[(X + Y E(X + Y )) 2 ] = E[((X E(X))) + (Y E(Y ))) 2 ] = E[( X + Ỹ )2 ] = E[ X 2 + Ỹ XỸ ] = E[((X E(X)) 2 ] + E[(Y E(Y )) 2 ] + 2E[((X E(X))(Y E(Y ))] = Var(X) + Var(Y ) + 2Cov(X, Y ) by definition of variance and covariance Hiro Kasahara (UBC) Econ / 28

18 Clicker Question 4-4 For any function g(x) and h(y), which of the following is true? A). Var(g(X) + h(y )) = Var(g(X)) + Var(h(Y )) B). Var(g(X) + h(y )) = Var(g(X)) + Var(h(Y )) + 2Cov(g(X), h(y )) C). Var(g(X) + h(y )) = g(var(x)) + h(var(y )) Hiro Kasahara (UBC) Econ / 28

19 Question 5 in Worksheet Prove that, for any constant a and b, Var[aX + by ] = a 2 Var[X] + b 2 Var[Y ] + 2abCov[X, Y ]. Hiro Kasahara (UBC) Econ / 28

20 Bernoulli random variable The probability mass function: { 0 with probability p X = 1 with probability 1 p What is E[X] and Var[X]? Hiro Kasahara (UBC) Econ / 28

21 Proposition If X is a Bernoulli random variable with P(X = 1) = p, then E[X] = p. Proof: E(X) def = x=0,1 xp x (1 p) 1 x = (0)(1 p) + (1)(p) = p Hiro Kasahara (UBC) Econ / 28

22 Clicker Question 4-6 Suppose that X 1 and X 2 are two Bernoulli random variables that are stochastically independent with P(X 1 = 1) = P(X 2 = 1) = p. Which of the following is true? A). E[X 1 + X 2 ] = p. B). E[X 1 + X 2 ] = 2p. C). E[X 1 + X 2 ] = p/2. Hiro Kasahara (UBC) Econ / 28

23 Question 7 in Worksheet Prove that, if X and Y are two Bernoulli random variables that are stochastically independent with P(X = 1) = P(Y = 1) = p, then E[X + Y ] = 2p. Hiro Kasahara (UBC) Econ / 28

24 Proposition If X is a Bernoulli random variable with P(X = 1) = p, then V [X] = p(1 p). Proof: Var(X) def = x=0,1 (x p) 2 p x (1 p) 1 x = (0 p) 2 (1 p) + (1 p) 2 p = p 2 (1 p) + (1 p) 2 p = (p + (1 p)) p(1 p) = p(1 p). Hiro Kasahara (UBC) Econ / 28

25 Clicker Question 4-7 Suppose that X 1 and X 2 are two Bernoulli random variables that are stochastically independent with P(X i = 1) for i = 1, 2. Which of the following is true? A). Var[(X 1 + X 2 )/2] = p(1 p). B). Var[(X 1 + X 2 )/2] = p(1 p)/2. C). Var[(X 1 + X 2 )/2] = p(1 p)/4. Hiro Kasahara (UBC) Econ / 28

26 Question 9 in Worksheet Prove that, if X and Y are two Bernoulli random variables that are stochastically independent with P(X = 1) = P(Y = 1) = p, then [ ] X + Y p(1 p) Var =. 2 2 Hiro Kasahara (UBC) Econ / 28

27 Proposition Suppose that X i for i = 1,..., n are n Bernoulli random variables that are stochastically independent to each other with P(X i = 1) for i = 1,..., n. Let X := 1 n n X i. Then, (1) E[ X] = p and (2) Var[ X] = p(1 p) n. Proof for (1): E[ X] = 1 n E[X 1 + X X n ] = 1 n {E[X 1] + E[X 2 ] E[X n ]} = p + p p n = n p = p. n Hiro Kasahara (UBC) Econ / 28

28 Proof for (2) Var[ X] = p(1 p) Var[ X] = Var = [ 1 n ] n X i n : ( ) 2 1 Var[X 1 + X X n ] n = 1 n 2 {Var[X 1] + Var[X 2 ] Var[X n ] +2Cov(X 1, X 2 ) + 2Cov(X 1, X 3 ) Cov(X n 1, X n )} = 1 {p(1 p) + p(1 p) p(1 p) + 0} n2 n p(1 p) = = n 2 p(1 p). n Hiro Kasahara (UBC) Econ / 28