Scribd
Upload
449 views

Polya Urn Model

The document describes Polya's urn model, which models repeated events that become more likely as previous occurrences of the event increase. It provides the mathematical formula for calculating the probability of drawing k red balls from an urn initially containing θ red and m white balls after n draws, where each draw is replaced along with an additional ball of the same color. The derivation of this formula is then shown through mathematical induction.

Uploaded by

Parmen
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
449 views

Polya Urn Model

The document describes Polya's urn model, which models repeated events that become more likely as previous occurrences of the event increase. It provides the mathematical formula for calculating the probability of drawing k red balls from an urn initially containing θ red and m white balls after n draws, where each draw is replaced along with an additional ball of the same color. The derivation of this formula is then shown through mathematical induction.

Uploaded by

Parmen
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 2

Derivation of Polya’s Urn Model

The Polya urn model1 is useful in survival analysis situations in which a system undergoes
repeated stresses, but the stresses are not independent of each other. For example, in the study of
the spread of a contagious disease in a closed population, the more infected people there are in
the population, the larger the likelihood that a non-infected person will become infected; i.e., the
more shocks affecting the individual. Polya’s urn model has often been applied to the spread of
contagious diseases.2

Consider an urn initially containing  red balls and m white balls. We randomly draw one ball
from the urn, note its color, and replace the ball, along with another ball of the same color. We
repeat this process n times. We want to know the probability distribution for the count of the
number of red balls drawn from the urn in n trials.

Proposition: Let Xn = number of red balls drawn in n draws. Then


P X n  k  
n!   k  1!  m  n  k  1!   m  1!
k! n  k !   1!  m  1!   m  n  1! , for k = 0, 1, 2, …, n.
Proof:
For n = 1, we have
P X n  0 
1!   0  1!  m  1  0  1!   m  1!  m
0!1  0 !   1!  m  1!   m  1  1!   m , the probability of
selecting a white ball from the urn on the initial draw; and
P  X n  1 
1!   1  1!  m  1  1  1!   m  1!  
1!1  1!   1!  m  1!   m  1  1!   m , the probability of
selecting a red ball from the urn on the initial draw.

Assume that the result holds true for n. Now we draw one more ball from the urn. The
probability that, after the (n+1)st draw, we will have drawn a total of k red balls is given by

 
P X n1  k   P red on  n  1 draw | X n  k  1 P X n  k  1
st

 P white on  n  1 
draw | X n  k P X n  k  .
st

Now, if k – 1 red balls were drawn in n draws, then at the (n+1)st draw, the urn contains
 + k – 1 red balls, and m + n – k + 1 white balls, so that the probability of obtaining a red ball is
  k 1
. On the other hand, if k red balls were drawn on the first n draws, then at the (n+1)st
 mn
draw, the urn contains  + k red balls and m + n – k white balls, so that the probability of
mnk
obtaining a white ball is .
 mn
Hence, the probability that k red balls are drawn in n + 1 draws is
  k 1 mnk
P X n 1  k   P X n  k  1  P X n  k  .
 mn  mn

By the induction hypothesis, we have


P X n  k  
n!   k  1!  m  n  k  1!   m  1!
k! n  k !   1!  m  1!   m  n  1! , and
P X n  k  1 
n!   k  2!  m  n  k !   m  1!
 k  1! n  k  1!   1!  m  1!   m  n  1! .
Therefore,
   k 1 
P X n 1  k   
n!   k  2!  m  n  k !   m  1!

   m  n   k  1! n  k  1!   1!  m  1!   m  n  1!

mnk  n!   k  1!  m  n  k  1!   m  1!

   m  n  k! n  k !   1!  m  1!   m  n  1! .

Collecting common factors and simplifying leads to


P X n 1  k  
 n  1!   k  1!  m  n  k !   m  1!
k! n  k  1!    1!  m  1!   m  n ! . #

The model may also be written as

n  k   knm   m


PXn k 
k  m  m n
, for k = 0, 1, 2, …, n.

1
Eggenberger, F. and Polya, G. (1923). Über die Statistik verketteter Vorgänge. Zeitschrift für
Angewandte Mathematik and Mechanik 1, 297-289.
2
Lundberg, O. (1964). On Random Processes and Their Application to Sickness and Accident
Statistics, 2nd Edition, Almqvist and Wiksells, Stockholm.

You might also like