An Introduction To Probability Theory and Its Applications II
An Introduction To Probability Theory and Its Applications II
1} occurs iff no disk center is con-
tained within 4, but it is known in advance that the circle of radius p
about the origin is empty. The remaining domain has area 2pt and we
conclude that the distribution of the visibility TL. is exponential:
PUL >t} = ett,4 WAITING TIME PARADOXES, THE POISSON PROCESS uw
In space the same argument applies and the relevant region is formed by
rotating our A about the z-axis, The rectangle 0
‘The break in the formula (4.3) at x = 1 is due to the special role of the
origin as the starting epoch of the process. Obviously
4.6) limo) = atze,
which shows that the special role of the origin wears out, and for an “old”
process the distribution of L, is nearly independent of 1. One expresses
this conveniently by saying that the “steady state” density of L, is given
by the right side in (4.6).
With the notations of the proof, the waiting time W, considered in the
‘example is the random variable W, = S,— {The argument of the proof
shows also that
Gan PO Saat tens & [eatntes-™ = ete dy
t-en
Thus W, has the same exponential distribution as the X, in accordance
with the reasoning (a). (See problem 7.)
Finally, a word about the Poisson process. The Poisson variables N(t)
were introduced as functions on the sample space of the infinite sequence
of random variables X;, X,,-... This procedure is satisfactory for many
purposes, but a different sample space is more natural. The conceptual
experiment “observing the number of incoming calls up to epoch 1” yields
for each positive ¢ an integer, and the result is therefore a step function with
unit jumps. The appropriate sample space has these step functions as sample
points; the sample space is a function space—the space of all conceivable
“paths.” In this space N(s) is defined as the value of the ordinate at epoch
tand S, asthe coordinate of the nth jump, etc. Events can now be considered
that are not easily expressible in terms of the original variables X,.. A typical
example of practical interest (see the ruin problem in VI,5) is the event that
N() > a+ bt for some f. The individual path (just as the individual
infinite sequence of 1 in binomial trials) represents the natural and un-
avoidable object of probabilistic inquiry. Once one gets used to the new
phraseology, the space of paths becomes most intuitive.Ls THE PERSISTENCE OF BAD LUCK 15
Unfortunately the introduction of probabilities in spaces of sample
paths is far from simple. By comparison, the step from discrete sample
spaces to the line, plane, etc., and even to infinite sequences of random
variables, is neither conceptually nor technically difficult. Problems of a
new type arise in connection with function spaces, and the reader is warned
that we shall not deal with them in this volume. We shall be satisfied with
an honest treatment of sample spaces of sequences (denumerably many
coordinate variables). Reference to stochastic processes in general, and to
the Poisson process in particular, will be made freely, but only to provide
an intuitive background or to enhance interest in our problems.
Poisson Ensembles of Points
As shown in 1; VI,6, the Poisson law governs not only “points dis-
tributed randomly along the time axis,” but also ensembles of points (such
as flaws in materials or raisins in a cake) distributed randomly in plane or
space, provided 1 is interpreted as area or volume. The basic assumption
was that the probability of finding k points in a specified domain depends
only on the area or volume of the domain, but not on its shape, and that
occurrences in non-overlapping domains are independent. In example
3(b) we used the same assumption to show that the probability that a domain
of volume 1 be empty is given by e~*', This corresponds to the exponential
distribution for the waiting time for the first event, and we see now that the
Poisson distribution for the number of events is a simple consequence of it,
‘The same argument applies to random ensembles of points in space, and we
have thus a new proof for the fact that the number of points of the ensemble
contained in a given domain is a Poisson variable. Easy formal calculations
may lead to interesting results concerning such random ensembles of points,
but the remarks about the Poisson process apply equally to Poisson en-
sembles; a complete probabilistic description is complex and beyond the
scope of the present volume.
5, THE PERSISTENCE OF BAD LUCK
As everyone knows, he who joins a waiting line is sure to wait for an
abnormally long time, and similar bad luck follows us on all occasions.
How much can probability theory contribute towards an explanation?
For a partial answer we consider three examples typical of a variety of
situations, They illustrate unexpected general features of chance fluctuations.
Examples. (a) Record values. Denote by Xo my waiting time (or financial
loss) at some chance event. Suppose that friends of mine expose themselves
to the same type of experience, and denote the results by Xj, Xs...
To exclude bias we assume that Xo,X,,... are mutually independent16 ‘THE EXPONENTIAL AND THE UNIFORM DENSITIES Ls
random variables with a common distribution, The nature of the latter
really does not matter but, since the exponential distribution serves as a
model for randomness, we assume the X, exponentially distributed in
accordance with (3.1). For simplicity of description we treat the sequence
{X)} as infinite.
To find a measure for my ill luck I ask how long it will take before a
friend experiences worse luck (we neglect the event of probability zero that
X, = X,). More formally, we introduce the waiting time N as the value of
the first subscript n such that X,>X.. The event {N>n—I} occurs
iff the maximal term of the n-tuple Xo, Xi,-..,Xq-1 appears at the initial
place; for reasons of symmetry the probability of this event is , The
event {N= n} is the same as (N>n—1}—{N>n}, and hence for
n=l,
G1) P(N =n} =
1d 1
ntl nati)’
This result fully confirms that I have indeed very bad luck: The random
variable N has infinite expectation! It would be bad enough if it took on the
average 1000 trials to beat the record of my ill luck, but the actual waiting
time has infinite expectation.
It will be noted that the argument does not depend on the condition that
the X, are exponentially distributed, It follows that whenever the variables
X, are independent and have a common continuous distribution function
F the first record value has the distribution (5.1). ‘The fact that this
distribution is independent of F is used by statisticians for tests of independ-
ence. (See also problems 8-11.)
The striking and general nature of the result (5.1) combined with the
simplicity of the proof are apt to arouse suspicion. The argument is really
impeccable (except for the informal presentation), but those who prefer to
rely on brute calculation can easily verify the truth of (5.1) from the direct,
definition of the probability in question as the (n+1)-tuple integral of
atienstzettzn) over the region defined by the inequalities 0 <2) <2,
and 0
6. WAITING TIMES AND ORDER STATISTICS
‘An ordered n-tuple (x,..., 24) of real numbers, may be reordered in
increasing order of magnitude to obtain the new n-tuple
ays Taps s+ sm) Where Ray S Fy S07 * S toy18 ‘THE EXPONENTIAL AND THE UNIFORM DENSITIES 16
This operation applied to all points of the space ‘R* induces n well-defined
functions, which will be denoted by Xqy,..-, Xia» If probabilities are
defined in 5" these functions become random variables. We say that
(Xay,--++Xiq)_ is obtained by reordering (X,,...,X,) according to
increasing magnitude, The variable Xi) is called kth-order statistic® of
the given sample X,,...,X,. In particular, Xj) and X,,) are the sample
extremes; when n = 2y +1 is odd, X,,41) is the sample median.
We apply this notion to the particular case of independent random
variables X,...,X, with the common exponential density ae~*.
Examples. (a) Parallel waiting lines. Interpret X,,...,X, as the lengths
of n service times commencing at epoch 0 at a post office with m counters.
The order statistics represent the successive epochs of terminations or, as
‘one might say, the epochs of the successive discharges (the “output process”).
In particular, Xq) is the waiting time for the first discharge. Now if the
assumed lack of aftereffect is meaningful, the waiting time Xj, must have
the Markov property, that is, Xq) must be exponentially distributed. As
a matter of fact, the event {Xy) > 1} is the simultaneous realization of
the n events (X, > 1}, each of which has probability e~*'; because of the
assumed independence the probabilities multiply and we have indeed
(6.1) P{Xq) > em
We can now proceed a step further and consider the situation at epoch
Xi). The assumed lack of memory seems to imply that the original situation
is restored except that now only m—1 counters are in operation; the
continuation of the process should be independent of Xi) and a replica of the
whole process. In particular, the waiting time for the next discharge,
namely Xi) ~ Xa), should have the distribution
(6.2) P{X)—Xay >t}
analogous to (6.1). This reasoning leads to the following general proposition
concerning the order statistics for independent variables with a common
exponential distribution.
"Strictly speaking the term “sample statistic" is synonymous with “function of the
sample variables,” that is, with random variable, It is used to emphasize linguistically the
different role played in a given context by the primary variable (the sample) and some
derived variables. For example, the “sample mean” (X,+"+X,)/n is called a statistic.
‘Order statistics occur frequently in the statistical literature, We conform to the standard
terminology except that the extremes are usually called extreme “values.”16 WAITING TIMES AND ORDER STATISTICS 19
Proposition* The variables Xqy),Xy — Xay +
independent and the density of Xs) — Xy is given by (n—K)ae™P™,
Before verifying this proposition formally let us consider its implications.
When n= 2 the difference X,) — Xi is the residual waiting time after the
expiration of the shorter of two waiting times. The proposition asserts that
this residual waiting time has the same exponential distribution as the original
waiting time and is independent of Xq). This is an extension of the Markov
property enunciated for fixed epochs t to the chance-dependent stopping
time Xq). It is called the strong Markov property. (As we are dealing with
only finitely many variables we are in a position to derive the strong Markov
property from the weak one, but in more complicated stochastic processes the
distinction is essential.)
‘The proof of the proposition serves as an example of formal mai
with integrals. For typographical simplicity we let m= 3. As in many
similar situations we use a symmetry argument. With probability one, no
two among the variables X; are equal, Neglecting an event of probability
zero the six possible orderings of X;, Xz, Xz according to magnitude there-
fore represent six mutually exclusive events of equal probability. To cal-
culate the distribution of the order statistics it suffices therefore to consider
the contingency X, < Xz < Xs. Thus
63) PAX) > fy Xer—Xay > fo» Xe Xe > bs} =
” = OP{X, > ty, Xe—Xy > tay Xp—Xp > 15}.
(Purely analytically, the space ‘R* is partitioned into six parts congruent to the
region defined by 2% <_< ty, each contributing the same amount to the
integral. The boundaries where two or more coordinates are equal have
probability zero and play no role.) To evaluate the right side in (6.3) we
have to integrate ae"**#**9) over the region defined by the inequalities
Phy AM >h Bom > hy
A simple integration with respect to. x leads to
emer an sede =
(64)
= tenn fara, = ementem
* This proposition has been discovered repeatedly for purposes of statistical estimation
but the usual proofs are computational instead of appealing to the Markov property. See
also problem 13,20 ‘THE EXPONENTIAL AND THE UNIFORM DENSITIES. 16
Thus the joint distribution of the three variables X,), Xi—Xavs Xay—Xewy
isa product of three exponential distributions, and this proves the proposition.
It follows in particular that E(Xoy—Xis) = (aK). Summing over
k=0,1,...,9-1 we obtain
(6.5) EX)
i
+o)
Note that this expectation was calculated without knowledge of the distri-
bution of X,, and we have here another example of the advantage to be
derived from the representation of a random variable as a sum of other
variables. (See 1; 1X,3.)
(8) Use of the strong Markov property. For picturesque language suppose
that at epoch 0 three persons 4, B, and C arrive at a post office and find
two counters free, The three service times are independent random variables
X, Y, Z with the same exponential distribution. The service times of
and B commence immediately, but that of C starts at the epoch Xa)
when either A or B is discharged. We show that the Markov property
leads to simple answers to various questions.
(i) What is the probability that C will not be the last to leave the post
office? ‘The answer is }, because epoch Xq) of the first departure establishes
symmetry between C and the other person being served.
(ii) What is the distribution of the time T spent by C at the post office?
Clearly T = Xq) + Z is the sum of two independent variables whose
distributions are exponential with parameters 2a and a. The convolution
of two exponential distributions is given by (2.14), and itis seen that T has
density u(t) = 2a(e*! — e-**) and E(T) = 3/(22)
(iii) What is the distribution of the epoch of the Jast departure? Denote
the epochs of the successive departures by Xa), Xia), Xie The difference
Xia) — Xqy is the sum of the two variables Xia) — Xie) and Xia) — Xo
We saw in the preceding example that these variables are independent and
have exponential distributions with parameters 2a and a. It follows that
Xia) — Xa) has the same density uw as the variable T. Now Xq) is
independent of Xj» Xa) and has density 2ae-*'. The convolution
formula used in (ji) shows therefore that X,,) has density
a
Aafe-t!—e-*!—are-tt]
and E(Xiq)) = 2/a.
The advantage of this method becomes clear on comparison with direct
calculations, but the latter apply to arbitrary service time distributions
(problem 19).
(©) Distribution of order statistics. As a final exercise we derive the
distribution of X,). The event {Xq) <1} signifies that at least k among47 THE UNIFORM. DISTRIBUTION 2
the m variables X, are 1 — 4 and by (7.1) the probability
for this equals 1°. The variable % has therefore density 2r, as asserted.
(Beginners are advised to try a direct computational verification.)
(@) Distribution of order statistics. If X,,...,%q are independent and
distributed uniformly in 0,1, the number of variables satisfying the in-
equality 0
You might also like