EC404 - Winter 2023 Archana Aggarwal

Introduction to Statistics and Econometrics 307, SSS-II

Answer Key to Problem Set 3

1. (a) Let X be the number of children in a randomly selected family. then E(X) = (0)(0.01) +
(1)(0.04) + (2)(0.03) + (3)(0.01) + (4)(0.01) = 1.7.
(b) A child picked at random is more likely to be in a large family than a small one. So, the
probability distribution for the number of children in the family that includes the randomly selected child
must place more weight on larger families. This is accomplished by weighting the probability law for the
number of children in a randomly selected family by the number of children in the family. Let W be the
number of children in the family of the randomly selected child.

Using this probability law, we obtain the expected number of children in the family of the randomly
selected child as

2. (a).

(b).

Page 1 of 9
Therefore, Var[Y]=E[Y2]!{E[Y]}2=7!(25'4)=0.75. You could also find the p.d.f. of Y and then calculate
its expectation and variance.

3. (a). Let G represent the event that the weather is good. Let B represent the event that the weather
is bad. We are given P(G) = P(B) = ½. To find the p.d.f of X, we first find the p.d.f of W, since
X = s + W = 2 +W. We know that given good weather, W 1 N(0, 1). We also know that given bad
weather, W 1 N(0, 4). The unconditional p.d.f of W is given by:

Now using the transformation X=2+W, we have

(b). P(1#X#3) = P(1#2+W#3) = P(!1#W#1) = P(!1#W #1) = P(G)P(!1#W#1|G) + P(B)P(!1#W #1|B)

= P(G)P (!1#W#1|G) + P(B)P

=P(G){P(W #1|G)!P(W #!1|G)} + P(B)

4. We have P(C|A) = 1!P(B), P(A) = k×P(B), and P(A1B) = P(A)×P(B) by independence.

6 as k $1 > P(A) > 0.

5. This is finite for all r. Therefore X has

moments of every order.

6. f(x+1) = (e -ëë(x+1))'[(x+1)!]= ë(e -ëë x)'[(x+1)(x!)]= [ë f(x)]'(x + 1), x = 0, 1, 2, 3, … , 8 > 0.

Page 2 of 9
7. (i)

(ii)

(iii) The probability that the random variable belongs to the interval (0,1'k) = F(1'k)= [1!(5' 2)e-1].

8. FX(x) = Pr(X # x) = pPr(X1 #x) + (1!p)Pr(X2 #x) = pF1(x) + (1!p)F2(x).

Page 3 of 9
9.

Notice how the properties of the exponential distribution result in the wage rate and the unemployment
rate being inversely related. Notice also that the regression of W on X is non-linear in this case.

10. X 1, X 2 and X 3 are independent random variables with X 1 1N(1,6), X 2 1N (11,2) and X 3 1 N (2,3). If
Z = X 1 +3 X 2 + 2X 3. Then:
(a) The distribution of Z is N (38, 36).
(b) P(Z > 0) = P[(Z-38)'6 >!38'6]=1.

11. Chebychev’s inequality implies that

Hence, Yn converges in probability to zero.

12. The distribution of X is N(6.05, 0.0004).

(a) P[(X!6.05 )'0.02 < (6.0171!6.05 )'0.02] = P[(X!6.05)'0.02< !0.0329'0.02)] = 0.05.
(b) The probability that at most two boxes (in a sample of size 9) weigh less than 6.0171 kg. each =
0.9916388 (using binomial with n = 9 and p = 0.05).
(c) Let be the sample mean of the 9 boxes.
P[( !6.05)'0.0067 #(6.035!6.05 )'0.0067] = 0.0125.

13. The distribution of Y is chi-squared with 18 d.f. The mean of Y is 18 and the variance of Y is 36.
P[(Y!18)'6#(9.39 !18)'6] = 0.07565 and P[(Y!18)'6 #(34.8!18)'6] = 0.9974. (Use the standard
normal tables to find these probabilities). Note that these values are larger than the exact values obtained
from the chi-squared tables.

Page 4 of 9
14. (a)

(b) The FOCs for a minimum of E[Y!(á+âX)]2 give us

Therefore, is the linear function of X that minimizes E[Y!(á+âX)]2.

15. Work with Zi = (X i!50)'5 (this is a standard normal variate) and use the CLT, i.e., find
P[Zi > (75!50)'5] from the standard normal tables. Here we don’t need the tables as P[Zi > 5] = 0.

16. P(Xi > 4) = c'4.

P(at most 10 sample values are > 4) = = the exact probability.

Let us set Y = the no. of Xi 's having values > 4. Then Y 1 Bin(12, 9) i.e., [(Y!12 )'3] 1 N(0,1), by the
CLT and P(0 #Y #10) . P(-4 #z #!2'3) = 1!0.747!0 = 0.253.

Page 5 of 9
17. (a) The probability that in a given show at most 200 seats are filled, given that the a ticket is

priced at Rs.150 is given by F(0.5) =

(b) The expected profit from a show, given a ticket price of Rs.150 = 150[E(Y)](400)-10,000 = 35,000.
(c) If ticket price is Rs.100 the new level of expected profit = 100[E(Y)](400) - 10,000 = 50,000'3.
Note that f(y) and hence E(Y) change as ticket price changes.

18. Expected profit = [P(winning the auction at price y)](1-y) = [P(y$yi, i=1,1,3,...,n-1)](1-y).

Maximizing with respect to y we get: á(n!1)y [á(n-1)-1] - [á(n-1)+1]yá(n!1) = 0, or

y = [á(n !1)]'[1+á(n-1)], from the first order condition. Checking the second order condition we find
that the above value will correspond to a point of expected profit maximization. The probability that he
will win the auction = yá(n-1). He should increase his bid as the number of bidders increases and his
expected profit will decline as the number of bidders increases.
Notice that the limiting value of y as n 6 4 is 1. Clearly as the number of bidders becomes large there is
no profit to be made.

19. For any two random variables X and Y, it is an identity that Y = a + bX + Y - a - bX for any
constants a and b that you may choose. Now I am interested in the particular values of a and b that satisfy
this condition: E(Y|X = x) = a + bX.
Take the conditional expectation of both sides of the identity above:
E(Y*X ) = E(a|X ) + E(bX |X ) + E (Y!a !bX |X ) = a + bX + E(Y!a!bX |X ).
Combining this with the condition E(Y|X ) = a + bX, we get E(Y!a!bX |X ) = 0. This tells us that
Cov(Y!a!bX, X)=0, because E[X(Y!a!bX)]=Ex{E[X(Y!a!bX)|X]}=E x[XE(Y!a!bX)|X)] = (E x[X])(0)=0.
Take the covariance of both sides of the identity with respect to X : Y = a+bX+Y!a!bX
6 Cov(Y,X ) = Cov(a+bX+(Y!a!bX), X)
6 Cov(Y,X ) = Cov(a, X) + Cov(bX, X) + Cov(Y!a!bX, X)
6 Cov(Y,X ) = 0 + bVar(X) + 0
6 b = Cov(Y,X )'Var(X).
Note: This result is important because it is the population version of the sample regression coefficient
used in econometrics. In other words, the most basic regression just estimates b by estimating Cov(Y,X )
and Var(X). Not all variables are amenable to a nice linear relationship like this. We have shown that if
there exist a and b satisfying E (Y *X = x) = a + bX, then b = Cov(Y,X )' Var(X), but often there will not
exist such a linear relationship. One case in which such a relationship always exists is if X and Y both
have Normal distributions.

20. The random variables Yi, are not independent. If Y1 and Y2 are independent then
Var(V) = Var(Y1) + Var(Y2).

Page 6 of 9
In this case , and hence Var(Y1) + Var(Y2) =

We compute

As Var(V) Var(Y1) + Var(Y2), Yi are not independent. Intuitively this is apparent from the fact that
consecutive Yi depend on the same value of Xi.
(b). Yes, they are identically distributed. Each Yi is the weighted sum of identical variables.
(c). E(Mn) = n-1E(Y1+Y2+.....+Yn) = E(Yi) = ì.

Var(Mn) =

21. Find the probability density of the random variable X, where the c.d.f. of X has the form:

The p.d.f. of X is obtained by taking the derivative of the c.d.f. of X: f(x) = dF(x)'dx = 1, 0 # x # 1.

22. The range of Y is from zero to one as 0#F(x)#1. The c.d.f. of Y is given by
G(y) = Probability (Y# y) = Probability (F(x) # y) = Probability (x # F -1(y)) = F(F -1(y)) = y.
From this we have the p.d.f. of Y as g(y) = dG(y)'dy = 1, 0 # y # 1.

23. , !1# x # 1. Therefore, the c.d.f of X =

Using the method followed in exercises 21 and 22 we have:

Range of Y is from zero to one, the c.d.f. of Y is
G(y) = P((X3+1)'2) # y) = P(X # (2y-1)a) = F(((2y-1)a) = (2y-1 + 1)'2 = y,
and the p.d.f. of Y is g(y) = dG(y)'dy = 1, 0 # y # 1, which is the p.d.f. of a uniform r.v..

24. The joint density function of W = X 2 and Z = Y 2 is obtained as follows:

First, G(w,z) = Probability(W#w, Z#z) = Probability(X 2#w, Y 2#z)
= Probability(-w½#X#w½,-z½#Y#z½) = F(w½,z½)-F(-w½,z½)-F(w½,-z½)+F(-w½,-z½).
Then, taking derivatives we have, g(w,z) = *J**[f(w½,z½)+f(-w½,z½)+f(w½,-z½)+f(-w½,-z½)]

Page 7 of 9
=
If X and Y are independent then, G(w,z) = Probability(W#w, Z#z) = Probability(X2#w, Y2#z)
= Probability(-w½#X#w½,-z½#Y#z½) = Probability(-w½#X#w½)*Probability(-z½#Y#z½)
= Fx(-w½#X#w½)*Fy(-z½#Y#z½) = Fx(w½)Fy(z½)-Fx(-w½)Fy(z½)-Fx(w½)Fy(-z½)+Fx(-w½)Fy(-z½);
and, g(w,z)=(4w½z½)-1*[fx(w½)fy(z½)+fx(-w½)fy(z½)+fx(w½)fy(-z½)+fx(-w½)fy(-z½)]

Thus, if X and Y are independent then W and Z are independent.

25. P(contamination) = 0.1, P(lithography) = 0.35, P(both) = 0.08. The chance that the chip will
not be successfully created is P(contamination) + P(lithography) ! P(both) = 0.37.
(a) The chance that the chip will be successfully created is 1!0.37 = 0.63.
(b) X ³ Bin(n=500, p=0.75), E(X) = np = 375, V(X) = np(1! p) = 93.75.
(c) P(X $ 365)=1!P(X < 365)=1!P(Z < (365!375)'9.68)=1!P(Z < !1.03)=1!0.1515 = 0.8485.

26. (a) Let wages be denoted by Y. Pr(Rs.10.80#Y#Rs.12.40)=Pr(Y#Rs.12.40)!Pr(Rs.10.80#Y).

Transform Y into a standard normal random variable by subtracting the population mean and dividing by

the population standard deviation. That is,

Here, ì = Rs.11.90 and ó = Rs.0.40, therefore, Pr(Rs.10.80 # Y # Rs.12.40)

= Pr[Z #(Rs.12.40 ! Rs.11.90)'Rs.0.40] !Pr[Z #(Rs.10.80 !Rs.11.90)'Rs.0.40]
= Pr(Z# 1.25) !Pr(Z #!2.75) = 0.8944 !0.0030 = 0.8914
Given many workers, the percentage of workers earning between Rs.10.80 and Rs.12.40 is 89 percent.

b) Pr(Y < Rs.11.00) = Pr[Z < (Rs.11.00 !Rs.11.90)'Rs.0.40] = Pr(Z < !2.25) = Pr(Z # !2.25)
= Pr(Z # 0) !Pr(!2.25 # Z # 0) = 0.5 !Pr(!2.25 #Z #0) = 0.5 !Pr(0 # Z # 2.25) = 0.5 !0.4878= 0.0122.
The percentage of workers earning less than Rs.11.00 is 1 percent.

c) Pr(Y >Rs.12.95) = Pr[Z > (Rs.12.95 !Rs.11.90)'Rs.0.40] = Pr(Z >2.625) = 1! Pr(Z # 2.625)
= 1! [Pr(Z # 0) + Pr(0 # Z # 2.625)] = 0.5 !0.4956 = 0.0044.
The percentage of workers earning more than Rs.12.95 is approximately 0.44 percent.

d) We want the wage a such that Pr(a # Y) = 0.10 = Pr{[(a ! Rs.11.90)'Rs.0.40] # Z}

= 0.5 !Pr{0 # Z #[(a ! Rs.11.90)'Rs.0.40]} = 0.1 6 Pr{0#Z #[(a !Rs.11.90)'Rs.0.40]}= 0.5!0.1 = 0.4.
Use the standard normal table to find c where Pr(0 #Z # c)=0.4 for c=[(a !Rs.11.90)'Rs.0.40].
c =1.28 = [(a !Rs.11.90)'Rs.0.40] 6 a = Rs.12.41. That is 10% of all workers earn more than Rs.12.41.

Page 8 of 9
e) We want the wage a such that Pr(Y # a) = 0.25 = Pr{Z #[(a ! Rs.11.90)'Rs.0.40]}. Note that a is
less than the population mean wage because only 25 percent of workers earn less than a, so
[(a ! Rs.11.90)'Rs.0.40] is negative. Therefore, Pr(Y # a) = Pr{Z #[(a ! Rs.11.90)'Rs.0.40]}
= Pr(Z #0) !Pr{[(a ! Rs.11.90)'Rs.0.40]#Z#0}= 0.5!Pr{[(a !Rs.11.90)'Rs.0.40] #Z #0} = 0.25.
6 Pr{0 # Z #[(a ! Rs.11.90)'Rs.0.40]} = 0.25.
Use the Z table to find c where Pr(Z # c) = 0.25 where c = ![(a ! Rs.11.90)'Rs.0.40].
c = 0.67 = ![(a ! Rs.11.90)'Rs.0.40] 6 a = Rs.11.63. That means that 25 percent of all workers earn
more than Rs.11.63.

f) We want the wages a and b such that Pr(Y # a) = 0.25 and Pr(Y # b) = 0.75. From part (e), we know
that a is Rs.11.63. So, all we have to compute is Pr(Y#b) = 0.75 or Pr(Y$b)= 0.25; as the normal p.d.f. is
symmetric around its population mean. Therefore, the points a and b are equidistant from the population
mean as Pr(Y # a) = 0.25 and Pr(Y $b) = 0.25. Thus, ì!a = b!ì. So, Rs.11.90 !a = b !Rs.11.90
6b = Rs.11.90 + Rs.11.90 !a = (Rs.11.90)(2)!Rs.11.63= Rs.12.17. The inter-quartile wage range, i.e.,
the wage range for the middle 50% of workers, is [Rs.11.63, Rs.12.17].

Page 9 of 9