Mark Osegard, Ben Speidel, Megan Silberhorn, and Dickens Nyabuti Discrete: Conditional Probability Mass Function P{X = x | Y = y} = P{X = x, Y = y} P{Y = y} Continuous: Conditional Probability Density Function f ( x, y ) fX | Y ( x | y ) := fX ( y ) Discrete: E[X | Y = y ] = xP{X = x | Y = y} x Continuous: E[X | Y = y ] = +∞ xfX | Y ( x, y )dx −∞ y E[X | Y = y ] of y. We write this as i.e. is a function E[X | Y ] E [ X | Y ]( y ) = E [X | Y = y ] (Conditional Expectation Function) E [X ] = E [E [X | Y ]] Clearly, when Y is discrete, E [X | Y = y ]P{Y = y} = y When Y is continuous, = +∞ −∞ E [X | Y = y ] fY ( y )dy Recall, if X,Y are jointly continuous with joint pdf f ( x, y ) Define: and f ( x, y ) fX | Y ( X | Y ) = fY ( y ) E[X | Y = y ] = +∞ −∞ xfX | Y ( X | Y )dx +∞ E [X | Y = y ] fY ( y )dy = −∞ +∞ +∞ xfX | Y ( x | y )dx fY ( y )dy −∞ −∞ = +∞+∞ xfX | Y (x | y ) f ( y )dxdy Y − ∞− ∞ f ( x, y ) x = fY ( y )dxdy fY ( y ) − ∞− ∞ +∞ +∞ = +∞ +∞ xf ( x, y )dxdy = − ∞− ∞ +∞ +∞ xf (x, y )dydx − ∞− ∞ (Fubini’s Theorem) = +∞ +∞ +∞ −∞ −∞ −∞ x f ( x, y )dydx = f ( x, y )dy +∞ xfX ( x )dx = E [X ] −∞ Therefore, concluding E [X ] = E [E [X | Y ]] When Y is discrete, E [X | Y = y ]P{Y = y} = y When Y is continuous, = +∞ −∞ E [X | Y = y ] fY ( y )dy [ Var ( X | Y ) = E ( X − E [X | Y ]) | Y [ 2 ] ] = E X | Y − (E [X | Y ]) 2 2 Var ( X ) = E [Var ( X | Y )] + Var (E [X | Y ]) E [X ] = E [E [X | Y ]] [ ] Var ( X | Y ) = E X | Y − (E [ X | Y ]) 2 2 [[ ] = E [E [X | Y ]]− E [(E [X | Y ]) ] = E [X ]− E [(E [X | Y ]) ] E [Var ( X | Y )] = E E X | Y − (E [X | Y ]) 2 2 2 2 2 2 ] ! " [( )] Var (E [ X | Y ]) = E E [X | Y ] − (E [E [ X | Y ]]) [ 2 2 ] = E (E [X | Y ]) − (E [X ]) 2 2 " [ ] E [Var ( X | Y )] + Var (E [ X | Y ]) = E X − (E [X ]) = Var ( X ) 2 Var ( X ) = E [Var ( X | Y )] + Var (E [ X | Y ]) 2 Definition Application to Probability Applications of Stopping Times to other formulas Basic Definition: A Stopping Time for a process does exactly that, it tells the process when to stop. Ex) while ( x != 4 ) { … } The stopping time for this code fragment would be the instance where x does equal 4. # Define: Suppose we have a sequence of Random Variables (all independent of each other) Our sequence then would be: X 1 , X 2 , X 3 ,... $ From our previous slide we have the sequence: X 1 , X 2 , X 3 ,... A discrete Random Variable N is a stopping time for this sequence if : {N=n} Where n is independent of all following items in the sequence X n +1 , X n + 2 ,... % Summarizing the idea of stopping times with Random Variables we see that the decision made to stop the sequence at Random Variable N depends solely on the values of the sequence X 1 , X 2 ,..., X n Because of this, we then can see that N is independent of all remaining values Xm,m > n $ Does Stopping Times affect expectation? No! Consider this statement: S= N i =1 Xi E[ S ] = E[ N ]E[ X ] This formula, the formula used for Conditional Expectation does remain unchanged $ For an example of how to use stopping times to solve a problem, we will now introduce to you Wald’s Equation… E[ N i =1 X i ] = E[ N ]E[ X ] & ' # If {X1, X2, X3, …} are independent identically distributed (iid) random variables having a finite expectation E[X], and N is a stopping time for the sequence having finite expectation E[N], then: E N i =1 X i = E[ N ]E[ X ] & ' Let N1 = N represent the stopping time for the sequence {X1, X2, …, XN } 1 Let N2 = the stopping time for the sequence {X(N +1) , X(N +2), …, X(N +N )} 1 1 1 2 Let N3 = the stopping time for the sequence {X(N +N +1) , X(N +N +2), …, X(N +N +N ) } 1 2 1 2 1 2 3 & ' We can now define the sequence of stopping times as {N i } i ≥ 1 where {Ni} clearly represents, {N1 , N 2 , N 3 , ...} and see the sequence is iid & ' " If we define a sequence {Si} as, where, {S i } = {S 1+ S 2+..... + S m } {S i } = S1 = N1 i =1 Note: {Si} are iid X i , S2 = N1 + N 2 i = N1 +1 X i , ... , Sm = N1 + N 2 +...+ N m Xi i = N1 + N 2 +...+ N m−1 & ' " {S i } = {S 1+ S 2+..... + S m } {N1+ N2+ ... + Nm} which are iid because the Xi’s are. with common mean E [Xi ] = E [X] & ' " By the Strong Law of Large Numbers, lim m →∞ {S1+ S2 + ... + Sm} E[X] = {N1+ N2+ ... + Nm} & ' " E[X ] Also {S 1+ S 2+..... + S m } {N 1+ N 2+..... + N m } m ↓ E [S ] m let m → ∞ ↓ E[N ] So as we let m→∞ E [ X ] = E [S ] E [N ] Which can be manipulated into our preposition: E[ S ] = E N i =1 X i = E[ N ]E[ X ] ( Sample Conditional and Stopping times in probability problem A miner is trapped in a room containing three doors. Door one leads to a tunnel that returns to the same room after 4 days; door two leads to a tunnel that returns to the same room after 7 days; door three leads to freedom after a 3 day journey. If the miner is at all times equally likely to choose any of the doors, find the expected value and the variance of the time it takes the miner to become free Using Wald’s Equation: X i ∈ {4,7,3} N= N = min{i | X i = 3} " , " # " ! 1 ∴ E [N ] = = 3 1 3 ) ( E [X ] = E [N ]E [ X i ] ' 14 E [X ] = 3 ∗ = 14 days 3 $ %& & Var ( X ) = E [Var ( X | N )] + Var (E [X | N ]) ( " E[X | N = n] = E = n i =1 n i =1 E[X i | N = n] Xi | N = n " E [X i | N = n] = xP{X i = x | N = n} = 4 P{X i = 4 | N = n}+ 7 P{X i = 7 | N = n} + 3P{X i = 3 | N = n} ∴ E [X i | N = n] = 11 i<n 2 3 i =n " n i =1 E[X i | N = n] = n −1 i =1 E [X i | N = n] + E [X i | N = n] 11 = (n − 1) + 3 2 11 5 = n− 2 2 11 5 ∴ E[X | N ] = N − 2 2 " Var (aX + b ) = a 2 Var ( x ) Var (E [ X 121 = Var 4 |N ]) = Var 11 5 N − 2 2 (N ) 1 1− 1− P 3 = 6 = Var ( N ) = 2 P2 1 3 Var ( X ) = E [Var ( X | N )] + Var (E [ X | N ]) || || ? 363 2 " Var ( X | N ) = Var n = i =1 n i =1 Xi | N = n Var ( X i | N = n ) [ ] Var ( X i | N = n ) = E X | N = n − (E [X i | N = n]) [ ] E X |N =n = 2 i 2 i x P{X i = x | N = n} 2 2 " = 4 2 P{X i = 4 | N = n}+ 7 2 P{X i = 7 | N = n} + 32 P{X i = 3 | N = n} 65 i<n 2 E Xi | N = n = 2 9 i=n [ ] 65 121 9 − = i<n Var ( X i | N = n ) = 2 4 4 9−9 = 0 i=n " Var ( X | N = n ) = n −1 i =1 Var ( X i | N = n ) + Var ( X n | N = n ) 9 = ( N − 1) 4 9 E [Var ( X | N )] = E ( N − 1) 4 9 9 9 = (E [N ] − E [1]) = (3 − 1) = 4 4 2 % Var ( X ) = E [Var ( X | N )] + Var (E [ X | N ]) || || 9 2 363 2 ) 9 363 372 Var ( X ) = + = = 186 2 2 2 * Dr. Steve Deckelman Dr. Ayub Hossain Who helped make this a success!!!!!!!!!!!!!!!!!!!!!!!