Scribd
Upload
6 views3 pages

Problem Sheet 3.3

This document contains 10 problems related to stochastic models and applications. The problems cover topics like probability of error in communication systems, finding probabilities and densities of transformations of random variables, algorithms for generating random variables with specific distributions, and methods for simulating discrete random variables.

Uploaded by

Niladri Dutta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views3 pages

Problem Sheet 3.3

This document contains 10 problems related to stochastic models and applications. The problems cover topics like probability of error in communication systems, finding probabilities and densities of transformations of random variables, algorithms for generating random variables with specific distributions, and methods for simulating discrete random variables.

Uploaded by

Niladri Dutta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

E1 222 Stochastic Models and Applications

Problem Sheet 3.3


(You need not submit solutions)

1. Consider a communication system. Let Y denote the bit sent by trans-


miter. (Y is a binary random variable). The receiver makes a measure-
ment, X, and based on its value decides what is sent. The decision at
the receiver can be represented by a function h : ℜ → {0, 1}. For any
specific h, let R0 (h) represent the set of all x ∈ ℜ for which h(x) = 0
and let R1 (h) represent the set of x ∈ ℜ for which h(x) = 1. An error
occurs if a wrong decision is made. Argue that the event of error is:
[h(X) = 0, Y = 1] ∪ [h(X) = 1, Y = 0]. Show that probability of error
for a decision rule h is
Z Z
p1 fX|Y (x|1)dx + p0 fX|Y (x|0)dx
R0 (h) R1 (h)

where pi = P [Y = i]. Now consider a h given by


h(x) = 1 if fY |X (1|x) ≥ fY |X (0|x)
(Otherwise h(x) = 0). Show that this h would achieve minimum prob-
ability of error.
2. Let X, Y have a joint distribution that is uniform over the quadrilat-
eral with vertices at (−1, 0), (1, 0), (0, −1) and (0, 1). Find P [X > Y ].
Are X, Y independent? (Hint: Can you decide on independence here
without calculating the marginal densities?)
3. Let X, Y be iid uniform over (0, 1). Let Z = max(X, Y ) and W =
min(X, Y ). Find the density of Z − W .
4. Let X, Y be iid exponential random variables with mean 1. Let Z =
X + Y and W = X − Y . Find the conditional density fW |Z
5. Let X, Y be independent Gaussian random variables with mean zero
and variance unity. Define random variables D and θ by

D = X 2 + Y 2; θ = tan−1 (Y /X)
(where, by convention, we assume θ takes values in [0, 2π]; for this
we first calculate tan−1 (|Y |/|X|) in the range [0, π/2] and then put

1
that angle in the appropriate quadrant based on signs of Y and X).
Find the joint density of D and θ and their marginal densities. Are D
and θ independent? (Hint: Note that the (D, θ) to (X, Y ) mapping is
invertible. Hence you can use the formula).

6. Consider the following algorithm for generating random numbers X


and Y :

1. Generate U1 and U2 uniform over [0, 1].


q q
2. Set X = −2 log(U1 ) cos(2πU2 ) and Y = −2 log(U1 ) sin(2πU2 ).

What would be the joint distribution of X and Y ? (Hint: Recall that


when U is uniform over [0, 1], we know that −a log(U ) is exponential
with parameter 1/a and 2πU is uniform over [0, 2π]).

7. Consider the following algorithm for generating random variables V1


and V2 :

1. Generate X1 and X2 uniform over [−1, 1].


2. If X12 + X22 > 1 then go to step 1; else set V1 = X1 , V2 = X2 and
exit.

What would be the joint distribution of V1 and V2 ?

8. Suppose we have access to a random number generator that can gener-


ate random numbers uniformly distributed over (0, 1). Using the results
of the previous problems, suggest a method for generating samples of
X when X has Gaussian density with mean zero and variance unity.

9. Let pi , qi , i = 1, · · · , N , be positive numbers such that N


i=1 pi =
P
PN
i=1 qi = 1 and pi ≤ Cqi , ∀i for some positive constant C. Consider
the following algorithm to simulate a random variable, X:

1. Generate a random number Y such that P [Y = j] = qj , j =


1, · · · , N . (That is, the mass function of Y is fY (j) = qj ).
2. Generate U uniform over [0, 1].
3. Suppose the value generated for Y in step-1 is j. If U < (pj /Cqj ),
then set X = Y and exit; else go to step-1.

2
On any iteration of the above algorithm, if condition in step-3 be-
comes true, we say the generated Y is accepted. Find the value of
P [Y is accepted | Y = j]. Show that P [Y is accepted, Y = j] = pj /C.
Now calculate P [Y is accepted]. Use these to calculate the mass func-
tion of X.

10. Suppose X is a discrete rv taking values {x1 , x2 , · · · , xm } with probabil-


ities p1 , · · · pm . The usual method of simulating such a rv is as follows.
We divide the [0, 1] interval into bins of length p1 , p2 etc. Then we
generate a rv, uniform over [0, 1] and depending on the bin it falls in,
we decide on the value for X. That is, if U ≤ p1 we assign X = x1 ; if
p1 < U ≤ p1 + p2 then we assgn X = x2 and so on.
Suppose X is a discrete random variable taking values 1, 2, · · · , 10. Its
mass function is: fX (1) = 0.08, fX (2) = 0.13, fX (3) = 0.07, fX (4) =
0.15, fX (5) = 0.1, fX (6) = 0.06, fX (7) = 0.11, fX (8) = 0.1, fX (9) =
0.1, fX (10) = 0.1. Can you use the result of previous problem to sug-
gest an efficient method for simulating X.

You might also like