This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 1 Great (Conditional) Expectations • X and Y are jointly discrete random variables • Recall, conditional expectation of X given Y = y: • Analogously, jointly continuous random variables: x Y X x y x p x y Y x X P x y Y X E )  ( )  ( ]  [  dx y x f x y Y X E Y X )  ( ]  [  Computing Probabilities by Conditioning • X = indicator variable for event A: E[X] = P(A) Similarly, E[X  Y = y] = P(A  Y = y) for any Y So: E[X] = E Y [ E X [X  Y] ] = E[ E[X  Y] ] = E[ P(A  Y) ] In discrete case: o Also holds analogously in continuous case Generalize, defining indicator variable F i = (Y = y i ): o Called “Law of total probability” otherwise occurs if 1 A X ) ( ) ( )  ( ] [ A P y Y P y Y A P X E y n i i i F P F A P A P 1 ) ( )  ( ) ( Hiring Software Engineers • Interviewing n software engineer candidates All n ! orderings equally likely, but only hiring 1 candidate o Claim: There is ato1 factor difference in productivity between the “best” and “average” software engineer o Depending on who you talk to, usually: 10 < a < 100 Right after each interview must decide hire/no hire Feedback from interview of candidate i is just relative ranking with respect to previous i – 1 candidates o Strategy: first interview k (of n ) candidates, then hire next candidate better than all of first k candidates o P k (best) = probability that best of all n candidates is hired o X = position of best candidate (1, 2, ..., n ) n i k n i k k i X P i X P i X P P n 1 1 )  Best ( ) ( )  Best ( ) Best ( 1 Hiring Software Engineers (cont.) Note: We will select best candidate (in position i ) if best of first i – 1 candidates is among the first k interviewed o To maximize, differentiate P k (Best) with respect to k : o Set g’ ( k ) = 0 and solve for k : o Interview n/e candidates, then pick best: P...
View
Full Document
 Spring '09
 Probability theory

Click to edit the document details