MSO205A PRACTICE PROBLEMS SET 12 SOLUTIONS

Question 1. Let X1 , X2 , X3 be a random sample from Bernoulli(p) distribution, for some p ∈ (0, 1).
Find the p.m.f. of X(2) .

Answer: X(2) is a discrete RV supported on {0, 1}. Now,

P(X(2) = 0)

= P(the second smallest of X1 , X2 , X3 is 0)

= P(at least two of X1 , X2 , X3 are 0)

= P(exactly two of X1 , X2 , X3 are 0)

+ P(all three of X1 , X2 , X3 are 0)

 
3
= (1 − p)2 p + (1 − p)3 , (using independence of X1 , X2 , X3 )
2
= (1 − p)2 (2p + 1).

3


Similarly, P(X(2) = 1) = 2 p2 (1 − p) + p3 = p2 (3 − 2p). Therefore, X(2) ∼ Bernoulli(p2 (3 − 2p)).

Question 2. Let X1 , · · · , Xn be a random sample from U nif orm(0, 1) distribution. Identify the
distribution of X(r) for r = 1, · · · , n.

Answer: We have FX(r) (x) = P(X(r) ≤ x) = 0, ∀x ≤ 0 and FX(r) (x) = P(X(r) ≤ x) = 1, ∀x ≥ 1.

For x ∈ (0, 1), P(Xi ≤ x) = x and P(Xi > x) = 1 − x, ∀i = 1, · · · , n. If Y is the number of
Xi , i = 1, · · · , n which fall in (0, x], then Y ∼ Binomial(n, x). Then,
n  
X n k
FX(r) (x) = P(X(r) ≤ x) = P(Y ≥ r) = x (1 − x)n−x .
k
k=r

In this case,
d n!
FX (x) = xr−1 (1 − x)n−r .
dx (r) (r − 1)!(n − r)!
Since FX(r) is differentiable everywhere except possibly at the points 0, 1, we conclude that X(r) is
a continuous RV with the p.d.f.

n! r−1
(r−1)!(n−r)! x (1 − x)n−r , if 0 < x < 1,


fX(r) =
0, otherwise.


and hence X(r) ∼ Beta(r, n − r + 1).

Question 3. Let X1 , · · · , Xn be a random sample from a distribution given by a p.d.f. f . Find the
joint p.d.f. of (X(r) , X(s) ) of 1 ≤ r < s ≤ n.
1
2 MSO205A PRACTICE PROBLEMS SET 12 SOLUTIONS

Answer: In the lecture notes, we have already discussed that the joint p.d.f. of (X(1) , · · · , X(n) ) is
given by

n! Qn

f (yi ), if y1 < · · · < yn ,
i=1
g(y1 , · · · , yn ) =
0, otherwise.


Now, integrating out co-ordinate variables yi , i ∈ 1, 2, · · · , n, i 6= r, i 6= s, we have


n! r−1
(r−1)!(s−r−1)!(n−s)! (F (yr )) [F (ys ) − F (yr )]s−r−1 (1 − F (ys ))n−s f (yr )f (ys ), if yr < ys ,


gX(r) ,X(s) (yr , ys ) =
0, otherwise


where F denotes the common DF for X1 , · · · , Xn .

Question 4. Let Y ∼ Np (b, K). Then for any c ∈ Rn and a n × p real matrix B, consider the n
dimensional random vector Z = c + BY . Show that Z ∼ Nn (c + Bb, BKB t ).

Answer: The Joint MGF of Y = (Y1 , Y2 , · · · , Yp ) is given by

 
1
MY (u) = exp ut b + ut Ku , ∀u ∈ Rp .
2

We compute the MGF of Z = (Z1 , · · · , Zn ). For any v ∈ Rn , we have

MZ (v) = E exp(v t Z) = E exp[v t (c + BY )]

= exp(v t c)E exp[(v t B)Y ]

= exp(v t c)MY (B t v)
 
t t 1 t t
= exp(v c) exp v Bb + v BKB v
2
 
1
= exp v t (c + Bb) + v t BKB t v .
2

Identifying the distribution of Z through the joint MGF, we conclude Z ∼ Nn (c + Bb, BKB t ).

Pp
Question 5. Let Y ∼ Np (b, K) with K being invertible. Then show that j=1 λj (Yj − bj ) = 0 for
some scalars λ1 , λ2 , · · · , λp if and only if λ1 = λ2 = · · · = λp = 0.

Answer: Any Variance-Coviance matrix is positive semi-definite. In our case, since K is invertible,
we conclude that K is positive definite. Then there exists an orthogonal matrix A such that
D = At KA is a diagonal matrix, with eigen-values of K as the diagonal entries.
If possible, let λt (Y − b) = 0 for some λ ∈ Rp . Then λt (Y − b)(Y − b)t λ = 0 and in particular,

λt Kλ = E λt (Y − b)(Y − b)t λ = 0.

Since K is positive definite, we conclude that λ = 0 ∈ Rp .

If λ = 0 ∈ Rp , we have λt (Y − b) = 0. This concludes the proof.
 MSO205A PRACTICE PROBLEMS SET 12 SOLUTIONS 3

P∞
Question 6. Let c := m=1 m−3 < ∞. Then the function f : R → [0, 1] given by

 1 x−3 , if x ∈ {1, 2, · · · }

c
f (x) =
0, otherwise


is a p.m.f.. Let X be a discrete RV with this p.m.f. and consider the following sequence of RVs
{Xn }n defined by 
X, if X ≤ n,

Xn = , ∀n.
0, otherwise


Show that the sequence of RVs {Xn }n converges in first mean to X, but not in the second mean.

Answer: We have, for any positive integer n

n n
1 X m 1 X 1
= < ∞,
c m=1 m3 c m=1 m2
and
n n
1 X m2 1 X 1
= < ∞.
c m=1 m3 c m=1 m
Therefore, EXn and EXn2 both exist, for all n. Moreover,
∞ ∞
1 X m 1 X 1 n→∞
E|Xn − X| = = −−−−→ 0,
c m=n+1 m3 c m=n+1 m2

but
∞ ∞
1 X m2 1 X 1
= = ∞, ∀n.
c m=n+1 m3 c m=n+1 m
Hence, {Xn }n converges to X in first mean, but not in the second mean.