Professional Documents
Culture Documents
Bayes Theorem in Decision Making
Bayes Theorem in Decision Making
Bayes Theorem in Decision Making
ABSTRACT Thomas Bayes (1702-1761) was an E n g l i s h m a t h e m a t i c i a n who d e v e l o p e d t h e f i r s t p r e c i s e , q u a n t i t a t i v e mathematical expression f o r i n d u c t i v e inference. H i s theorem on i n v e r s e p r o b a b i l i t y , p u b l i s h e d p o s t h u m o u s l y i n 1763, y i e l d s t h e p r o b a b i l i t y t h a t an e v e n t t h a t h a s a l r e a d y o c c u r r e d may have o c c u r r e d i n a p a r t i c u l a r way, and a f f o r d s a powerful tool f o r determining root cause. The e s s e n c e o f Bayes' Theorem i s t h a t the p r o b a b i l i t y o f o c c u r r e n c e o f one p a r t i c u l a r hypot h e s i s i s equal to i t s c o n d i t i o n a l p r o b a b i l i t y d i v i d e d by the sum o f the c o n d i t i o n a l p r o b a b i l i t i e s of a l l p o s s i b l e h y p o t h e s e s .
A l t h o u g h Bayes' Theorem e v o l v e d from p o s t u l a t e s o f c l a s s i c a l p r o b a b i l i t y , i t h a s been t h e s u b j e c t o f c o n t r o v e r s y because o f s u b j e c t i v i t y i n v i e w i n g p r i o r e v e n t s . There h a s n e v e r been a q u e s t i o n , however, about t h e m a t h e m a t i c a l v a l i d i t y of t h e theorem; where e m p i r i c a l d a t a o r e x p e r t o p i n i o n a r e a v a i l a b l e , t h e theorem p r o v i d e s r e a l world conclusions.
T h i s paper s t a r t s w i t h a r e v i e w of p r o b a b i l i t y fundamentals to ensure a f i r m f o o t i n g f o r unders t a n d i n g B a y e s ' Theorem. The u s e o f t h e theorem i s i l l u s t r a t e d w i t h i n c r e a s i n g l y complex a p p l i c a t i o n s to e n s u r e t h e r e a d e r ' s g r a s p o f i t s v e r s a t i l i t y and p o t e n t i a l , and t o p r o v i d e models f o r a p p l i c a t i o n s unique to t h e r e a d e r .
PROBABILITY FUNDAMENTALS
INTRODUCTION
The a p p l i c a t i o n s o f B a y e s ' Theorem o f I n v e r s e P r o b a b i l i t y r a n g e from m e d i c i n e , s c i e n c e , e n g i n e e r i n g , and m a n u f a c t u r i n g t h r o u g h e c o n o m i c s , p o l i t i c s , and m i l i t a r y s t r a t e g y . Judicious use of B a y e s ' Theorem g r e a t l y i m p r o v e s t h e q u a l i t y o f d e c i s i o n s , e s p e c i a l l y where r o o t c a u s e s o f d i s c e r n e d e f f e c t s a r e t o be deduced, o r where a s s o c i a t e d r i s k s o f a l t e r n a t i v e a p p r o a c h e s a r e t o be evaluated. B a y e s ' Theorem employs both a p r i o r i proba b i l i t y and a p o s t e r i o r i p r o b a b i l i t y to e v a l u a t e c a u s a l hypotheses. I t s c o n c l u s i o n s a r e both p r e c i s e m a t h e m a t i c a l l y and tempered by e m p i r i c a l d a t a . I t i s o n l y c o n t r o v e r s i a l because of the s u b j e c t i v e n a t u r e o f a s s i g n i n g a p r i o r i p r o b a b i l i t i e s to e v e n t s y e t to take p l a c e .
The p r o b a b i l i t y o f an event i s a number from 0 to 1. T h i s i n d i c a t e s the r e l a t i v e f r e q u e n c y w i t h which the event would o c c u r i f the t r i a l were r e peated a l a r g e number of t i m e s . From t h e mathem a t i c a l v i e w p o i n t , assignment o f p r o b a b i l i t i e s o f o c c u r r e n c e to e v e n t s i s a x i o m a t i c ; such a s s i g n m e n t s are a r b i t r a r i l y based on c e r t a i n immutable p o s t u lates. F o r example, the sum o f the p r o b a b i l i t i e s of o c c u r r e n c e o f m u t u a l l y e x c l u s i v e and c o l l e c t i v e l y e x h a u s t i v e e v e n t s must t o t a l 1. Such p o s t u l a t e s are t h e b a s e s f o r t h e d e f i n i t i o n s g i v e n i n T a b l e I , as w e l l as f o r the r u l e s f o r c a l c u l a t i n g p r o b a b i l ity values that follow.
Addition
Rule
where P ( E + F ) denotes the p r o b a b i l i t y o f o c c u r r e n c e of E o r F o r both, and P ( E F ) t h e j o i n t p r o b a b i l i t y of o c c u r r e n c e o f E and F . F o r example, i f P ( E ) = 0.6 and P ( F ) = 0.2, t h e n : P(E+F) = 0.6 + 0.2 - 0.12 = 0.68 exclusive events, (2) then: (3)
where E and F a r e m u t u a l l y
Copyright 1987 M A R T I N M A R I E T T A C O R P O R A T I O N Reprinted with permission of Martin Marietta Corporation. ,
P(EF) 44
=0
1987
SAVE CONFERENCE
PROCEEDINGS
where P(EF) denotes the j o i n t p r o b a b i l i t y o f occurrence o f E and F. Again, i f P(E) = 0.6 and P(F) = 0.2, then: P(EF) = ( 0 . 6 ) ( 0 . 2 ) = 0.12. Product Rule-Dependent Events The p r o b a b i l i t y o f occurrence of two dependent events i s given by:
P(EF) = P(E)P(F|E)
A P o s t e r i o r i P r o b a b i l i t y - P r o b a b i l i t y determined a f t e r t r i a l or experimental information i s obcained. Also called empirical p r o b a b i l i t y . A P r i o r i P r o b a b i l i t y - P r o b a b i l i t y determined be fore any t r i a l or experimental information i s obtained. Also called mathematical p r o b a b i l i t y . Conditional P r o b a b i l i t y - P r o b a b i l i t y of occurrence of an event Chat depends on the occurrence o f another event in the sample space.
(7)
(8)
P(FE)
P(F)P(E|F)
(9)
C o l l e c t i v e l y ExhausCive EvenCs - T o t a l i t y of a l l possible outcomes of a t r i a l . Compound Events - Events composed of a group o f mutually exclusive events, and whose p r o b a b i l i t y of occurrence i s the sum of the p r o b a b i l i t i e s of occurrence of Che mutually exclusive events. Dependent Events - Possible outcomes of a C r i a l whose i n d i v i d u a l p r o b a b i l i t i e s of occurrence depend on outcomes bf oCher events. Equally Probable Events - Possible outcomes of a t r i a l whose i n d i v i d u a l p r o b a b i l i t i e s of occurrence are equal. Independent Events - Possible outcomes of a t r i a l whose i n d i v i d u a l p r o b a b i l i t i e s of occurrence are independent of the occurrence of other events. J o i n t P r o b a b i l i t y - P r o b a b i l i t y of occurrence o f two or more events. Marginal P r o b a b i l i t y - A p r i o r i p r o b a b i l i t y of occurrence of dependenc events. Mutually Exclusive Events - Events whose occurrences preclude the occurrence of other events w i t h i n the same sample space (e.g., heads or tails). Sample Space - Set of c o l l e c t i v e l y exhaustive events. T r i a l - Endeavor having more than one possible outcome,
where P ( F | E ) denotes the c o n d i t i o n a l p r o b a b i l i t y o f F g i v e n the occurrence o f E and P ( E | F ) the condit i o n a l p r o b a b i l i t y o f E given the occurrence o f F .
Consider the random s e l e c t i o n o f three parts frora a l o t o f f i v e c o n t a i n i n g two d e f e c t i v e p a r t s , for example. What i s the j o i n t p r o b a b i l i t y o f a l l of the three selected parts being good? The j o i n t p r o b a b i l i t y o f the f i r s t three p a r t s selected being good i s the marginal p r o b a b i l i t y o f the f i r s t p a r t selected being good m u l t i p l i e d by the successive conditional p r o b a b i l i t i e s : P(GiG2G3) = P(Gi)P(G2|Gi)P(G3|GiG2). (10)
The marginal p r o b a b i l i t y o f the f i r s t part selected being good i s simply the number of good p a r t s d i v i d e d by the t o t a l number o f p a r t s , o r : P(Gi) (11)
The c o n d i t i o n a l p r o b a b i l i t y o f the second p a r t s e l e c t e d being good given the f i r s t p a r t selected i s good i s simply the r a t i o o f the remaining number of good p a r t s and nuraber o f p a r t s remaining, o r : P(G2lGi) = | . (12)
where P(E+F) now denotes the p r o b a b i l i t y o f occurrence o f E o r F, but not b o t h . I n t h i s case the example y i e l d s : P(E+F) = 0.6 + 0.2 = 0.8. Product Rule-Independent Events The j o i n t p r o b a b i l i t y o f occurrence o f two o r more independent events i s g i v e n by: P(EF) = P(E)P(F) (6) 45 (5)
10.
1987
events. I t s p e c i f i e s the p r o b a b i l i t y of o c c u r r e n c e o f an e v e n t t h a t c a n c o n d i t i o n a l l y o c c u r i n s e v e r a l ways and i s so c a l l e d b e c a u s e i t e l i m i n a t e s noncontributing factors. I t s e t s the stage f o r def i n i n g B a y e s ' Theorem.
P(E|H^) P(H^)P(E|H^)
H2
E_ a r e e v e n t s i n a sample I f E l , E2, s p a c e , t h e n t h e p r o b a b i l i t y o f o c c u r r e n c e o f an e v e n t F i n t h e sample s p a c e c a n be e x p r e s s e d by:
P(E|H2) - P(H2)P(E|H2)
p(F)
.1^
P(E-)-P(F|E^)
(15)
of F g i v e n
F o r example, the p r o b a b i l i t i e s o f E v e n t s E j , E2, and a r e 0,50, 0.30 and 0.10, and t h e p r o b a b i l i t i e s o f Event F , g i v e n t h e o c c u r r e n c e o f E v e n t s E j , E ^ , and E ^ , a r e 0.08, 0.05 and 0.02. A p p l y i n g E q u a t i o n 15 y i e l d s :
P(H.)
J
P(E|H.)
J
P(Hj)P(E|Hj)
Figure
1.
P(F)
BAYES' THEOREM
The a p p l i c a t i o n s t h a t f o l l o w a r e i n c r e a s i n g l y complex t o e n s u r e a f i r m g r a s p o f p r i n c i p l e s and a p p r e c i a t i o n o f t h e p o t e n t i a l o f f e r e d by B a y e s ' Theorem toward s o l v i n g r e a l - w o r l d p r o b l e m s . The r e a d e r i s urged t o c o n s i d e r t h e c o n c l u s i o n s one might r e a c h w i t h o u t t h e c a u s a l i n s i g h t t h e theorem provides.
B a y e s ' Theorem i s a l s o c a l l e d t h e Theorem o f I n v e r s e P r o b a b i l i t y because i t p r o v i d e s the proba b i l i t y t h a t an e v e n t t h a t h a s a l r e a d y o c c u r r e d m i g h t have o c c u r r e d i n a p a r t i c u l a r way, w h i c h i s tantamount t o r e a s o n i n g from e f f e c t t o c a u s e . S t a t e d more p r e c i s e l y , B a y e s ' Theorem p r o v i d e s p r o b a b i l i t i e s of occurrence f o r hypothetical causes by m o d i f y i n g a p r i o r i p r o b a b i l i t i e s f o r e v e n t s on the b a s i s o f t r i a l e v i d e n c e , e x p e r t o p i n i o n , o r other measures o f a p o s t e r i o r i p r o b a b i l i t y .
APPLICATIONS
Cost Estimating A government agency w i s h e s to p r o c u r e a new d a t a p r o c e s s i n g f a c i l i t y and needs t o know t h e most l i k e l y turn-key cost f o r budgeting. Data given i n T a b l e I I r e l a t e c o s t hypotheses to t h e i r r e s p e c t i v e p r o b a b i l i t i e s a s e s t i m a t e d by t h e a g e n c y ' s s t a f f and c o n c l u d e d from i n f o r m a l q u o t a t i o n s by p o t e n t i a l suppliers.
I t i s c o n v e n i e n t t o v i e w B a y e s ' Theorem i n t h e p r o b a b i l i t y t r e e form shown i n F i g u r e 1. The p r o b a b i l i t y t h a t E v e n t E was r e a c h e d by t h e j - t h b r a n c h o f t h e t r e e , g i v e n t h a t i t was r e a c h e d t h r o u g h one of i t s n branches, i s the r a t i o of the p r o b a b i l i t y a s s o c i a t e d w i t h t h e j - t h b r a n c h t o t h e sum o f t h e p r o b a b i l i t i e s a s s o c i a t e d w i t h a l l n branches of the tree. 46
$4,000,000 $3,500,000
lity
1987
what
P(5,OOO,000lQi) =
The a p r i o r i p r o b a b i l i t i e s o f u s i n g a component from the r e s p e c t i v e v e n d o r s , V^, a r e g i v e n as: (18) P ( V i ) = 0 . 4 5 , P(V2)=0.30, P(V3)=0.25. (22)
where P(5,000,000|Q^) d e n o t e s the p r o b a b i l i t y o f the f a c i l i t y c o s t i n g $5,000,000 g i v e n the s u p p l i e r quotations. The answer i s more c o n s e r v a t i v e than t h e most c o n s e r v a t i v e s t a f f o p i n i o n ( T a b l e I I ) .
Game o f Chance E a c h o f f o u r u r n s i s to be sampled once by the player. The f i r s t u r n c o n t a i n s one r e d b a l l and two w h i t e b a l l s , the second one red b a l l and t h r e e w h i t e b a l l s , the t h i r d one r e d b a l l and f o u r w h i t e b a l l s , and the f o u r t h one r e d b a l l and f i v e w h i t e balls. The s e l e c t i o n o f u r n s i s e q u a l l y p r o b a b l e .
prob-
ball
Pl
= f,
P2 = f
P3
5'
PA
i 6,
(19)
P(D|V2) =
(0.30)(0.03)/[(0.45)(0.06) +(0.30)(0.03)+(0.25)(0.02)]
(26)
=0.22
P(D|V3) =
(0.25)(0.02)/[(0.45)(0.06 +(0.30)(0.03)+(0.25)(0.02] .
(27)
=0.12
P(Ui)
A p p l y i n g E q u a t i o n 17 f o r P ( U 2 I R ) , w h i c h d e n o t e s the p r o b a b i l i t y o f h a v i n g s e l e c t e d U2 g i v e n a r e d b a l l was p i c k e d , y i e l d s :
P(U2lR)
15 57
(21)
D e f e c t i v e Components I d e n t i c a l components a r e p r o c u r e d from t h r e e v e n d o r s f o r use i n t h e n e x t h i g h e r a s s e m b l y o f a production process. F o r t y - f i v e p e r c e n t of the components a r e p r o c u r e d from Vendor 1, 30 p e r c e n t from Vendor 2, and 25 p e r c e n t from Vendor 3. The r e s p e c t i v e p e r c e n t a g e s o f d e f e c t i v e s a r e 6, 3, and 2.
P o l l u t i o n d e t e c t i o n d e v i c e s of the e n v i r o n m e n t a l p r o t e c t i o n agency of a c e r t a i n s t a t e c a n d e t e c t e x c e s s i v e amounts o f p o l l u t a n t s e m i t t e d by f a c t o r i e s w i t h a p r o b a b i l i t y o f 0.90, and p r o b a b i l i t y of 0.20 t h a t f a c t o r i e s not e x c e e d i n g limits w i l l f a i l the t e s t . The i s s u e i s whether to p r o c u r e d e v i c e s w i t h a d e t e c t i o n p r o b a b i l i t y o f 0.99, even though the i n c r e a s e d s e n s i t i v i t y w i l l i n c r e a s e the f a l s e a l a r m p r o b a b i l i t y to 0.22, to apprehend raore v i o l a t o r s o f s t a t e s t a t u t e s . I t i s assumed 30 p e r c e n t o f t h e f a c t o r i e s i n the s t a t e emit excessive pollutants.
F i g u r e 2 i l l u s t r a t e s the p r o b a b i l i t y t r e e a p p r o a c h to the problem. H y p o t h e s e s and H2 a r e the f r a c t i o n s o f v i o l a t o r s and n o n - v i o l a t o r s , and E v e n t E \ i s the s t a t e of d e t e c t i n g e x c e s s i v e e m i s s i o n s ; E2 i s the s t a t e of d e t e c t i n g none x c e s s i v e e m i s s i o n s and i n d i c a t i n g e x c e s s i v e ( i . e . , false alarm).
47
0.90 ( 0 . 3 0 ) ( 0 . 9 0 ) = 0.27
INITIAL CONDITIONS
0.99 ( 0 . 3 0 ) ( 0 . 9 9 ) = 0.297
( 0 . 7 0 ) ( 0 . 2 2 ) = 0.154
F i g u r e 2.
Environmental
P r o t e c t i o n Example
The f o l l o w i n g e x p r e s s i o n i s used to c a l c u l a t e the r e q u i r e d MTBF from s p e c i f i e d r e l i a b i l i t y and overhaul i n t e r v a l v a l u e s : p = g-t/MTBF (30) 2.718. 30 y i e l d s : (31)
e equals
= 0.66 .
(28)
= 0.66 .
(29)
The MTBF o f 9,950 hours s e r v e s as t h e nominal d e s i g n v a l u e and might have a t o l e r a n c e o f p l u s o r minus t e n p e r c e n t f o r v a l u e s o f 10,945 o r 8,955 hours. The q u e s t i o n i s one o f v a l i d a t i n g t h e s p e c i f i e d v a l u e o f MTBF w i t h o u t t e s t i n g e v e r y engine over the o v e r h a u l i n t e r v a l , which d e c r e a s e s overall life expectancy.
problem c o n s i s t s of the
Life
Expectancy C o n f i d e n c e f a c t o r s a r e a s s i g n e d to the MTBF t o l e r a n c e l i m i t s u s i n g knowledge o f p a r t s v a r i a b i l i t y , and e n g i n e e r i n g e x p e r i e n c e and judgment. The f a c t o r s a r e used as a p r i o r i p r o b a b i l i t i e s o f r e a l i z i n g the r e s p e c t i v e MTBF v a l u e s i n the d e s i g n , and a r e l i s t e d i n T a b l e I I I f o r t h e example.
A c l a s s i c a l example o f a p p l y i n g B a y e s ' Theorem concerns the l i f e expectancy of j e t engines. Typic a l l y , r e l i a b i l i t y g o a l s a r e e x p r e s s e d i n terms o f the r e q u i r e d p r o b a b i l i t y o f f a i l u r e - f r e e o p e r a t i o n over t h e time i n t e r v a l between s u c c e s s i v e e n g i n e overhauls.
2.
The a r e a under the c u r v e to the r i g h t of t i s the p r o b a b i l i t y t h a t t h e time between s u c c e s s i v e f a i l u r e s i s equal to o r g r e a t e r t h a n t f o r the g i v e n v a l u e o f mean time between f a i l u r e s (MTBF). S i m i l a r l y , t h e a r e a under t h e curve to t h e l e f t o f
The c o n d i t i o n a l p r o b a b i l i t i e s o f s u c c e s s f u l t r i a l a r e c a l c u l a t e d given t h a t the r e s p e c t i v e MTBF v a l u e s were r e a l i z e d i n the d e s i g n . T h i s step f i r s t r e q u i r e s the c a l c u l a t i o n of marginal p r o b a b i l i t i e s of f a i l u r e - f r e e o p e r a t i o n over time p e r i o d s e q u a l t o t h e r e s p e c t i v e MTBF v a l u e s .
48
1987
combinations
Probability
P(MTBF) 0.10 0.80 0.10 Next, m a r g i n a l p r o b a b i l i t i e s of f a i l u r e - f r e e o p e r a t i o n over time p e r i o d s e q u a l to t h e r e s p e c t i v e MTBF v a l u e s a r e c a l c u l a t e d . A p p l y i n g E q u a t i o n 30 yields: ^8,995 = e-8.995/9',950 = o.40 (34)
4.
B a y e s ' Theorem i s t h e n used t o c a l c u l a t e the a p o s t e r i o r i p r o b a b i l i t y o f r e a l i z i n g the s p e c i f i e d MTBF v a l u e i n t h e d e s i g n on the b a s i s o f t h e a p r i o r i p r o b a b i l i t i e s from S t e p 1, g i v e n t h e c o n d i t i o n a l p r o b a b i l i t i e s from Step 3 and a s u c c e s s f u l trial.
(35)
(36)
5.
These v a l u e s equal the a r e a under t h e c u r v e to the r i g h t o f t i n F i g u r e 3, when t e q u a l s t h e respective periods. I n a d d i t i o n , the a r e a s under the c u r v e to the l e f t o f t a r e the p r o b a b i l i t i e s o f f a i l u r e i n the r e s p e c t i v e p e r i o d s , t h a t i s ( 1 - P ) .
C o n t i n u i n g t h e example, assume t h e s u c c e e d f a i l r a t i o i s 5:2; t h a t i s , f i v e e n g i n e s a r e o p e r a t e d f o r 100 h o u r s each and two o r fewer f a i l u r e s c o n s t i t u t e s u c c e s s . Table IV gives the sample s p a c e f o r t h e s u c c e s s f u l t r i a l w i t h G and B c o n n o t i n g good o r bad r e s u l t s .
n-r
r
s (37)
(1-P)
r=o
G G G G G B G G
G G G G B G G G
G G G B G G G B
G G B G G G B G
G B G G G G B B
G B G G B G B B
B G G B G B G B
G G B G G B B G
G G B B B G G G
B B G G G G G G
where P ( E | M T B F ) d e n o t e s the c o n d i t i o n a l proba b i l i t i e s of Event E, i n t h i s case a c h i e v i n g the s p e c i f i e d s u c c e e d - f a i l r a t i o , g i v e n the r e s p e c t i v e MTBF v a l u e s , and s d e n o t e s t h e number o f s t a t e s i n t h e sample space f o r the r e s p e c t i v e number of failures r .
A p p l y i n g t h e v a l u e s frora E q u a t i o n s 34, 35, and 36, and number o f s t a t e s from T a b l e I V , i n E q u a t i o n 37 y i e l d s : 1. F o r MTBF o f 8,995 P ( E |8995)= hours:
(38)
2.
F o r MTBF o f 9,950
hours:
C?
, r!
^, (n-r)!
(32)
(39)
1987
3.
F o r MTBF o f 10,945
hours:
CONCLUSION Bayes' Theorem i s a t t h e h e a r t o f modern s t a t i s t i c a l d e c i s i o n theory. I t i s a powerful tool f o r r e a s o n i n g from c a u s e t o e f f e c t , and, c o n v e r s e l y , p r e d i c t i n g t h e outcomes o f a l t e r n a t i v e strategies.
(40)
The p r o b a b i l i t y o f t h e s p e c i f i e d MTBF v a l u e being r e a l i z e d i n the design, given that the s p e c i f i e d s u c c e e d - f a i l r a t i o i s a c h i e v e d , c a n now be c a l c u l a t e d by a p p l y i n g B a y e s ' Theorem a s e x p r e s s e d i n E q u a t i o n 17:
P(9950|E) =
(41)
T h i s v a l u e i s l e s s than t h e 0.8 c o n f i d e n c e f a c t o r r e q u i r e d by e n g i n e e r i n g , and even i f t h e s p e c i f i e d s u c c e e d - f a i l r a t i o were a c h i e v e d i n t h e t r i a l , t h e d e s i g n s h o u l d be deemed o n l y m a r g i n a l l y acceptable. The MTBF t o l e r a n c e l i m i t s about t h e d e s i g n g o a l s h o u l d be t i g h t e n e d .
50