Académique Documents
Professionnel Documents
Culture Documents
e2(mN/2) /N
P (m) = p
N/2
so P (270) = 7 103 (Can also use Stirlings approx.).
1
2
DN
2
so that < DN
>= N l2 . This immediately generalizes to random walk in d dimensions, provided one can walk in the direction of coordinate axes only.
Then we have
d
X
2
DN =
(D( i))2N .
i=1
where Di is the distance traveled in the i th direction. Now at every step we have
a choice of the direction in which to go, and we will assume that all of them are
equally probable, and of equal length l. there is 1/d probability to go in any one
of them, so:
2
<
2
DN
1X
[(Di )N 1 + l]2 + [(Di )N 1 l]2 X
d
2
2
>=
=1 <
+ (Dj )2N1 >=< DN
1 > +l
d i
2
j6=i
2
2
Also, < D1 >= l . Thus, the RMS deviation from the origin after N steps in
N l in any number of dimensions.
Problem 5 : Birthdays
a )
Let the probability that two or more people out of a group of N have the same
birthday be p. Then p = 1 q, where q is probability of all the people having
distinct birthdays.
We count the total number of possible events DN , where D = 365. The number
D!
of cases all people have distinct birthdays is (DN
)! , so
p=1
D!
DN .
(D N )!
b )
The task is to find N such that p(N ) 21 . Since:
p=1
D!
DN
(D N )!
2D ln 2 22.
It is clear that our approximation of D N being large was valid. If we now plug
N = 22 into the equation A above we get 0.48 for p. Thus we actually need a
larger N . Usine N = 23 gives p = 0.51. Thus 23 is the number of people for which
the probability of two or more coincident birthdays is 1/2.
3
c )
Whats the probability that 100 people all have different birthdays ?
Using the Stirlings approximation:
q(N = 100) (
D
)DN eN 3 107
DN
Problem 6
A neutrino detector obtains a small background signal of = 1count/day.
Then,after a T = 2 yrs long experiments the physicists observe a signal of 3 events
in one hour. Is this a statistical fluctuation? Let p be the probability of observing
at least 3 events in t = 1hr. Probability P that the rate exceeds two events in
one hour in the duration of experiment is:
P = 1 Q,
where Q is probability that the rate never does exceed 2/hr:
T
Q = (1 p) t .
Detection of background neutrinos at such a small rate is governed by a Poisson
distribution. Probability of obtaining k events in time t due to random fluctuations is given by
t
ke
P (k; t, ) = (t)
k!
Probability of obtaining at least rate k0 is
p = P (k0 ; t, ) =
ke
(t)
k!
kk0
=1
X
k<k0
ke
(t)
k!
Problem 7
We have here a binomial distribution with the probability p = 0.6 of success
(Clinton winning the election). If we sample 10 people, the probability that all of
them would vote for Bush is:
P10 = (1 p)10
4
Similarly the probability that 9 of the 10 people would vote for him is :
10
P9 =
p(1 p)9
9
etc. The probability that our poll will indicate that Bush will win is:
P =
k10
X
Pk
k>5
1623424
= 0.166239
9765625
Now to estimate the probability that a poll of 1000 people will indicate that
Bush will win The standard deviation for a binomial distributions is:
p
p
= N pq = N p(1 p) 15.5
For our poll to indicate that Bush will win, we would need 500+ votes for this
= 6.5 away from the expected value
candidate. Thats more than roughly 500400
15.5
of 400 votes.
We can roughly approximate the distribution of votes for Bush as a gaussian
with mean = 400 and standard deviation = 15.5 Then the probability of the
poll being in favor of Bush is 0 to much better than 1 part in 1010 :
Z +
(x)2
1
e 22 4.016 1011
2 2
+6.5