session 4
DESCRIPTION
TRANSCRIPT
7/3/2012
1
How reliable is Reliable?
Let D denote the event that an IC is defective
Let DR denote the event that ‘reliable’ test finds the IC to be defective
Given, P(D) = .005 P[DR| D] = 198/200 = 0.99
P [DR| not D] = 8/160 = 0.05
P[DR] = P[DR and D] + P[DR and not D]
= P[D] × P[DR| D] + P[not D] × P[DR| not D]
= 0.005 × 0.99 + 0.995 × 0.05 = 0.00495 +0.04975 = 0.0547
P[D| DR] = P[DR and D] / P[DR] = 0.00495/0.0547 = 0.0905
P[not D| DR] = 0.9095
Same calculation using table
Events Prior Prob
(i)
P[DR|..]
(ii)
P[DR and ..]
(iii)
=(i)*(ii)
P[..|DR]
(iv)
=(iii)/P[DR]
D 0.005 0.99 0.00495 P[D| DR]=0.0905
Not D 0.995 0.05 0.04975 P[not D| DR]=0.9095
Sum 1 P[DR]=
0.0547
If Reliable finds the IC to be
not defective
Events Prior Prob
(i)
P[not DR|..]
(ii)
P[not DRand ..]
(iii)
P[..|not DR]
D 0.005 0.01 0.00005 P[D| not DR]=0.000053
Not D 0.995 0.95 0.94525 P[not D|
not DR]=0.999947
Sum 1 P[not DR]=
0.9453
Bayes Rule
][][
]|[][]|[
][
][
cBP
BAP
ABPBAP
ABPAPBAP
∩+∩
×==
∩
]|[][]|[][
]|[][cc ABPAPABPAP
ABPAP
×+×
×=
Bayes Rule -- using table
events Prior Prob
P[Ai]
(i)
P[B|Ai]
(ii)
P[BAi]
(iii)
=(i)*(ii)
P[Ai|B]
(iv)
=(iii)/P[B]
A1 P[A1] P[B|A1] P[A1B] P[A1| B]
A2 P[A2] P[B|A2] P[A2B] P[A2| B]
Ak P[Ak] P[B|Ak] P[AkB] P[Ak| B]
Sum 1 P[B]
7/3/2012
2
First Can contains 10 marbles: 7 red & 3 blue
Second Can contains 4 blue & 1 red
1 marble 1 marble
1 marble
P(the marble drawn from the
Box is blue)=?
The marble drawn from
the Box is found to be
blue. What is the prob.
that it came from Can2?
Monty Hall Problem
or
‘Khul Ja SimSim’
• The car is behind one of the 3 doors.
• You select door A.
• Aman Verma (who knows where the car is)
opens door B and shows that this is empty.
Gives you an option of “switch”.
• Should you stick to your initial choice?
Solution to KJSS
Correct
Door
Prior
Probabi
lity
P(door B is opened
given you select
door A and correct
door is..)
P( correct door is /
and door B is
opened and you
selected door A)
P( correct door is /
given Door B is
opened and you
selected door A)
A 1/3 1/2 1/6 1/3
B 1/3 0 0 0
C 1/3 1 1/3 2/3
P(door B is
opened) 1/2
Will the new product do well?Combine prior opinion with pilot survey
Will do well = 40% of potential customers will like it
Won’t do well = 20% will like it
PRIOR PROB
60%
40%
Pilot survey: from 25 potential customers
X like it
Want P[will do well | X out of 25 like it]
AB
Is it easier to find
P[B|A] ?