You are on page 1of 8

Stats Hw 1

Jon Chen
January 12, 2016
Problem 1
Alternative 3 is not more likely than alternative 2. Assume the probability
of 1 is F where F is greater than 0 and less than 1, and that the probability
of 2 is B also greater than 0 and less than 1. Statement 3 is an intersection
of 1 and 2, making the probability of 3 to be the product of F and B. Since
both F and B are in between 0 and 1, the product must be less than both F
and B.
1. Let the probability that statement 1 is true be F such that
0<F <1

(1)

2. Let the probability that statement 2 is true be B such that


0<B<1

(2)

3. Let the probability that statement 3 is true be S such that


S =F B

(3)

S =F B

(4)

4. by the definitions in (1) and (2),


S F, S B

(5)

This is a fallacy because the answer that most people get to is false. This
may be explained that since only information that F is likely to happen
is given and nothing is said about B, when given a choice that contains
that information such as statement 3, people intuitive link it to being more
probable, even though logically it would not be.
1

Problem 2
The word given is CONSTANTINOPLE. It contains 14 letters in total. So,
if all the letters were unique, the total set of possible permutations is 14!.
However, we note that the word is not made up of all unique letters. Instead,
it has 2 O, 2 T, and 3 N. Thus the total number of permutations is now
14!
. So the probability of these 14 letters to be randomly arranged in
2! 2! 3!
14!
this specific order is 1 in
, or 2.75 1010 .
2! 2! 3!

Problem 3
The word given is SLUMGULLION. It contains 3 Ls, 4 vowels, and 4 consonants. To find the subset of arrangements that give a word with all the Ls
preceding the consonats, we first remove all the vowels. Now we force the
condition that all the Ls precede the consonants. since there are 4 unique
consonant, there are a total of 4! ways to arrange the letters in this condidtion. Now we seperate all the setters and return teh vowels to the word one
at a time. The first vowel would have a choice of 8 possible spaces to choose,
the next would have 9, the next 10, and the last 11. Thus the total number
of arrrangements for all the letters such that all the Ls precede the consants
is now 4! 8 9 10 11. However, we note that there are 2 Us, so to adjust
for that, we divide the total number of possibilities by 2!. Thus the total
4! 8 9 10 11
, or 95040.
number of possible arragment is
2

Problem 4
In this problem, the task is to essentially find out the total number of possible
paths from the top of the diamond to the bottom. Since we have to start
at the top, we imagine that there are 2 path choices at each node or row,
either L or R. Since there are a total of 11 rows, there will be a total of 10
choices. Also, since the end point is directly below the start point, we know
that of the 10 choices, 5 must be L and 5 mush be R. Thus the total possible
10!
number of paths will be
. The numerator of 10! gives us the total
5! 5!
possible number of paths given 10 choices, while the denominator removes
the double counting from the 5 indifferentiable Ls and Rs.

Problem 5
Let us call A, B and C the events in which A, B, and C are pardoned respectively. We start off with all three prisoners with the same probability of
being pardoned. Giving us:
P (A) = P (B) = P (C) =

1
3

(6)

Now, let us call b the case where the warden says that B is to be executed.
So, now we have established that
P (b|B) = 0

(7)

P (b|C) = 1

(8)

1
(9)
2
However, what we really want to know is given b, what is P(A), or P (A|b)
We can find P (A|b) by using bayes theorem
P (b|A) =

P (A|b) =

P (b|A)P (A)
P (b|A)P (A) + P (b|B)P (B) + P (b|C)P (C)

So we find
P (A|b) =

1
2

1
2

13
13 + 0 13 + 1

1
3

1
3

(10)

(11)

2
1
Thus the P(A) is still at whereas now P(C) is at So, prisoner C is correct
3
3
Extra
if a biased coin was used that names B with some probability p such that
1
p 6= , then the value of P (b|A) would be p. Thus:
2
p 13
P (A|b) =
p 13 + 0 13 + 1

1
3

p
p+1

(12)

Problem 6
Let us call P(A) the probability that a child is abused and P(N) the probability that he is not. From the problem, we are given that
P (A) =

1
90

(13)

89
(14)
90
Now, let us call a the event where the program says there is abuse and n the
event where it says there isntt. From the problem, we are give that
P (N ) =

P (a|A) =

9
10

(15)

3
(16)
100
Again, since we are only interested in the probability that a child is abused
given that the program identifies him, in other words P (A|a), we use Bayes
theorem to convert the given quantities to the desired result.
P (a|N ) =

P (A|a) =

P (a|A)P (A)
= 0.2521
P (a|A)P (A) + P (a|N )P (N )

(17)

So the probability of actual abuse given that the program says its is abuse is
only about 0.25.
If the number of incinence were changed, then the resulting P(A) and
P(N) would change in the equation given above. If incindences were 1 in
1
999
1000, P (A) = 1000
and P (N ) = 1000
, resulting in P (A|a) = 0.0292.
49
1
and P (N ) = 50
, resulting in P (A|a) =
If incindences were 1 in 50, P (A) = 50
0.3798

Problem 7
a) The probability that a path is clear is the intersection of p and q
P = P (p q) = p q

(18)

b) The probability that a path is clear is the union of p and q


P = P (p q) = p + q p q

(19)

c) For convienience, I will call the probability that the path is clear to be
the following: W C = p = 0.8, CT = q = 0.9, W D = p0 = 0.9, DT =
q0 = 0.8. First, we seperate the scenario into 2 cases, case A where CD
is not passable and case B where CD is. Looking only at case A, we
see that the probability that case A is clear is
P (A) = p q + p0 q0 p q p0 q0 = 0.92
4

(20)

Looking only at case B, we see that it is easier to find the probability


of not being able to pass. So we will find that first. Since the only
requirement for her to not pass is for p and p0 to be blocked or q and
q0 to be blocked. Or as shown below:
1 P (B) = (1 p)(1 p0) + (1 q)(1 q0) (1 p)(1 p0)(1 q)(1 q0)
(21)
So we find that P(B) = 0.96
However, the chance of A occuring is 0.05 and B is 0.95, thus we multiply the factors to each case respectively. Finally, the probability that
any path is passable is the union of 0.05*P(A) and 0.95*P(B), and since
A and B are mutually exclusive, the actual probability is
P = 0.05 P (A) + 0.95 P (B) = 0.96

Problem 8
For this problem, we split the probability up. So we find the probability of
winning given the result of the first throw. Thus the probability of winning
P (W ) is now a sum of the probabilities of the 7 possible initial win conditions,
which are the following:
1. winning with 7 or 11 on first roll P (I)
2. winning with 4 on the first roll P (4)
3. winning with 5 on first roll P (5)
4. winning with 6 on first roll P (6)
5. winning with 8 on first roll P (8)
6. winning with 9 on first roll P (9)
7. winning with 10 on first roll P (10)
We note that due to the symmetric nature of the 2 dice roll,
P (4) = P (10), P (5) = P (9), P (6) = P (8)
Now, we look first at the case of P(4). The initial probability of rolling a
3
. To win, the any roll onward just needs to be a 4. The
4 with 2 dice is 36
chances of rolling neither a 4 nor a 7, or in other words to move on the next
roll is 27
. Since P(4) is the sum of all the winning probability for each roll,
36
it can be written as:
3 3
27 3
27
3
27
3
(22)
P (4) = ( )[ + ( )( ) + ( )2 ( ) + ( )3 ( ) + ....]
36 36
36 36
36 36
36 36
5

or

X
3
27
P (4) =
( )2 ( )n
36 36
n=0

(23)

Using the geometric series, we find that


P (4) =

1
36

(24)

Using the same formula as shown above, we can calculate the probability of
winning for each first roll
Event P (event)
1
4
36
16
5
360
25
6
396
25
8
396
16
9
360
1
10
36
8
I
36
Thus the probability of winning craps is
P (W ) = P (I) + P (4) + P (5) + P (6) + P (8) + P (9) + P (10) = 0.493 (25)

Problem 9
To prove Bonferonis inequality:
\

n
n
X
P
Aj 1
[1 P (Aj )]
j=1

(26)

j=1

We first consider the case of n=2. We have already established that


P (A1 A2 ) = P (A1 ) + P (A2 ) P (A1 A2 )

(27)

Next, we can find the complement from each of the elements in the equality
above and establish the equality below:
P (A1 A2 )c = P (A1 )c + P (A2 )c P (A1 A2 )c

(28)

1 P (A1 A2 ) = [1 P (A1 )] + [1 P (A2 )] [1 P (A1 A2 )]

(29)

P (A1 A2 ) = 1 [1 P (A1 )] [1 P (A2 )] + [1 P (A1 A2 )]

(30)

Since by the defination of probability, we know that 0 [1P (A1 A2 )] 1.


Thus we can establish the following inequality:
P (A1 A2 ) 1 [1 P (A1 )] [1 P (A2 )]
6

(31)

Now, we can generalize this inequality if we show that it holds for k+1
elements. So again, we start with the following:
P (A1 A2 ...Ak Ak+1 ) =P (A1 ) + P (A2 ) + ...P (Ak ) + P (Ak+1 )
P (A1 A2 ...Ak Ak+1 )

(32)

We now take the complement of each of the elements in the equality above
P (A1 A2 ...Ak Ak+1 )c =P (A1 )c + P (A2 )c + ...P (Ak )c + P (Ak+1 )c
P (A1 A2 ...Ak Ak+1 )c
(33)
P (A1 A2 ...Ak Ak+1 ) =1 (1 P (A1 )) (1 P (A2 )) ...
(1 P (Ak )) (1 P (Ak+1 ))
(34)
+ [1 P (A1 A2 ...Ak Ak+1 )]
The equaiton can be condensed using sigma notation to the following:
P

k+1
\


Aj

=1

j=1

k+1
X

[1 P (Aj )] + [1 P (A1 A2 ...Ak Ak+1 )]

(35)

j=1

Again, since we know that 0 [1 P (A1 A2 ...Ak Ak+1 )] 1, the


equality can be rewiten as the following inequality:
P

k+1
\


Aj

j=1

k+1
X

[1 P (Aj )]

(36)

j=1

Problem 10
The minimal possible percentage of those who lost all is simply the minimum intersection between the 4 subsets. Using Bonfferonis inequality from
problem 9, we find that
P 1 [(1 0.7) + (1 0.75) + (1 0.8) + (1 0.85)] = 0.1

(37)

R excercise
dice <- function(x){ ## this functon is the simulated dice roll
roll <- sample(1:6, x, replace = T)
return (roll) }
numround <- function(x) { ## this function simulates the 24 rolls
record <- c()
## and returns 1 if a pair of aces was
state <- c(0)
## rolled and 0 if not for each round
i <- 0
while (i < x) { ## this loop rolls the 2 dice,
q <- dice(1)
## and records whether or not a pair of aces
p <- dice(1)
## was rolled
value <- q+p
record <- c(record,value)
i <- i+1
if (value == 2){
state <- c(1)
break
}
}
tally <- c(record, state)
return (state)
}
numgame <- function(x){ ## this function simulates the 1000 games
count<- c()
## that was played, and counts the number
i <- 0
## of rounds that won
while (i<x){
count <- c(count,numround(24))
i<- i+1
}
return (count)
}
prob<- sum(numgame(1000))/1000
print (prob)

## this line takes the number


## of wins and normalizes it to a
## probability of winning

You might also like