STAT2006B Assignment 1

Due Time and Date: 5 p.m., February 20, 2023

INSTRUCTION: Submit ONE file which includes the solutions of all questions. The file
must be in PDF format. You may write down your solutions on paper and scan them or
type up the solution.
Question 1
A random sample 𝑋𝑋1 , 𝑋𝑋2 , … , 𝑋𝑋𝑛𝑛 of size 𝑛𝑛 is taken from 𝑁𝑁(𝜇𝜇, 𝜎𝜎 2 ), where the variance 𝜃𝜃 = 𝜎𝜎 2 is such that 0 < 𝜃𝜃 <
∞ and 𝜇𝜇 is a known real number. Show that the maximum likelihood estimator for 𝜃𝜃 is 𝜃𝜃� = (1⁄𝑛𝑛) ∑𝑛𝑛𝑖𝑖=1(𝑋𝑋𝑖𝑖 − 𝜇𝜇)2
and that this estimator is an unbiased estimator of 𝜃𝜃.
Question 2
For determining half-lives of radioactive isotopes, it is important to know what the background radiation isotopes, it
is important to know what the background radiation is in a given detector over a specific period. The following data
were taken in a 𝛾𝛾-ray detection experiment over 98 ten-second intervals:

58 50 57 58 64 63 54 64 59 41 43 56 60 50
46 59 54 60 59 60 67 52 65 63 55 61 68 58
63 36 42 54 58 54 40 60 64 56 61 51 48 50
60 42 62 67 58 49 66 58 57 59 52 54 53 53
57 43 73 65 45 43 57 55 73 62 68 55 51 55
53 68 58 53 51 73 44 50 53 62 58 47 63 59
59 56 60 59 50 52 62 51 66 51 56 53 59 57

Assume that these data are observations of a Poisson random variable with mean 𝜆𝜆.
a) Find the values of 𝑥𝑥̅ and 𝑠𝑠 2 .
b) What is the value of the maximum likelihood estimator of 𝜆𝜆?
c) Is 𝑆𝑆 2 an unbiased estimator of 𝜆𝜆?
d) Which of 𝑥𝑥̅ and 𝑠𝑠 2 would you recommend for estimating 𝜆𝜆? Why? You could compare the variance of 𝑋𝑋� with
the variance of 𝑆𝑆 2 , which is
𝜆𝜆(2𝜆𝜆𝑛𝑛 + 𝑛𝑛 − 1)
𝑉𝑉𝑉𝑉𝑉𝑉(𝑆𝑆 2 ) = .
𝑛𝑛(𝑛𝑛 − 1)
Question 3

Let 𝑓𝑓(𝑥𝑥; 𝜃𝜃) = (1⁄𝜃𝜃)𝑥𝑥 (1−𝜃𝜃)⁄𝜃𝜃 , 0 < 𝑥𝑥 < 1, 0 < 𝜃𝜃 < ∞.

(a) Show that the maximum likelihood estimator of 𝜃𝜃 is 𝜃𝜃� = −(1⁄𝑛𝑛) ∑𝑛𝑛𝑖𝑖=1 ln 𝑋𝑋𝑖𝑖 .

(b) Show that 𝐸𝐸�𝜃𝜃�� = 𝜃𝜃 and thus that 𝜃𝜃� is an unbiased estimator of 𝜃𝜃.
Question 4
Let 𝑋𝑋1 , 𝑋𝑋2 , … , 𝑋𝑋𝑛𝑛 be a random sample from 𝑏𝑏(1, 𝑝𝑝) (i.e., 𝑛𝑛 Bernoulli trials). Thus,
 𝑛𝑛

𝑌𝑌 = � 𝑋𝑋𝑖𝑖 is 𝑏𝑏(𝑛𝑛, 𝑝𝑝).

𝑖𝑖=1

(a) Show that 𝑋𝑋� = 𝑌𝑌⁄𝑛𝑛 is an unbiased estimator of 𝑝𝑝.

(b) Show that 𝑉𝑉𝑉𝑉𝑉𝑉(𝑋𝑋�) = 𝑝𝑝(1 − 𝑝𝑝)/𝑛𝑛.

(c) Show that 𝐸𝐸[𝑋𝑋�(1 − 𝑋𝑋�)/𝑛𝑛] = (𝑛𝑛 − 1)[𝑝𝑝(1 − 𝑝𝑝)/𝑛𝑛2 ].

(d) Find the value of 𝑐𝑐 so that 𝑐𝑐𝑋𝑋�(1 − 𝑋𝑋�) is an unbiased estimator of 𝑉𝑉𝑉𝑉𝑉𝑉(𝑋𝑋�) = 𝑝𝑝(1 − 𝑝𝑝)/𝑛𝑛.
Question 5
Let 𝑋𝑋1 , 𝑋𝑋2 , … , 𝑋𝑋𝑛𝑛 be a random sample of size 𝑛𝑛 from a normal distribution.
(a) Given that the distribution of (𝑛𝑛 − 1)𝑆𝑆 2 /𝜎𝜎 2 is 𝜒𝜒 2 (𝑛𝑛 − 1), show that an unbiased estimator of 𝜎𝜎 is 𝑐𝑐𝑆𝑆, where
𝑛𝑛 − 1
√𝑛𝑛 − 1Γ � 2 �
𝑐𝑐 = 𝑛𝑛 .
√2Γ �2�
∞
HINT: Γ(𝑧𝑧) = ∫0 𝑡𝑡 𝑧𝑧−1 𝑒𝑒 −𝑡𝑡 𝑑𝑑𝑡𝑡.

(b) Find the value of 𝑐𝑐 when 𝑛𝑛 = 5; when 𝑛𝑛 = 6.

(c) Graph 𝑐𝑐 as a function of 𝑛𝑛. What is the limit of 𝑐𝑐 as 𝑛𝑛 increases without bound? Note: you may use any software
to produce the graph.
Question 6
In some situations where the regression model is useful, it is known that the mean of 𝑌𝑌 when 𝑋𝑋 = 0 is equal to 0,
i.e., 𝑌𝑌𝑖𝑖 = 𝛽𝛽𝑥𝑥𝑖𝑖 + 𝜀𝜀𝑖𝑖 where 𝜀𝜀𝑖𝑖 for 𝑖𝑖 = 1, 2, … , 𝑛𝑛 are independent and 𝑁𝑁(0, 𝜎𝜎 2 ). Obtain the maximum likelihood
estimators, 𝛽𝛽̂ and 𝜎𝜎 �2 , of 𝛽𝛽 and 𝜎𝜎 2 under this model.

Question 7

• 𝑝𝑝). We know that 𝑋𝑋� is an unbiased estimator of 𝑝𝑝 and that

Let 𝑋𝑋1 , 𝑋𝑋2 , … , 𝑋𝑋20 denote a random sample from 𝑏𝑏(2,
𝑉𝑉𝑉𝑉𝑉𝑉(𝑋𝑋�) = 𝑝𝑝(1 − 𝑝𝑝)/𝑛𝑛.
(a) Find the Rao-Cramér lower bound for the variance of every unbiased estimator of 𝑝𝑝.

(b) What is the efficiency of 𝑋𝑋� as an estimator of 𝑝𝑝?

Question 8

Find the Rao-Cramér lower bound, and thus the asymptotic variance of the maximum likelihood estimator 𝜃𝜃�, if the
random sample 𝑋𝑋1 , 𝑋𝑋2 , … , 𝑋𝑋𝑛𝑛 is taken from each of the distribution having the following pdfs:

(a) 𝑓𝑓(𝑥𝑥; 𝜃𝜃) = (1⁄𝜃𝜃 2 )𝑥𝑥𝑒𝑒 −𝑥𝑥⁄𝜃𝜃 , 0 < 𝑥𝑥 < ∞, 0 < 𝜃𝜃 < ∞.

(b) 𝑓𝑓(𝑥𝑥; 𝜃𝜃) = (1⁄(2𝜃𝜃 3 ))𝑥𝑥 2 𝑒𝑒 −𝑥𝑥⁄𝜃𝜃 , 0 < 𝑥𝑥 < ∞, 0 < 𝜃𝜃 < ∞.

(c) 𝑓𝑓(𝑥𝑥; 𝜃𝜃) = (1

A ⁄𝜃𝜃)𝑒𝑒 (1−𝜃𝜃)/𝜃𝜃
- , 0 < 𝑥𝑥 < 1, 0 < 𝜃𝜃 < ∞.

0×0-1
 4¥
PH -

_ ×) = ' o e-

PHKH.tw/y. Xn--Xn)-11¥
PHi=XdPHi- X4..]P(Xn--X/n)¥•é%÷
- #-e-
110 ) =

:#
*
é%÷
.
. .
.

* # exp # E. Him )
"

MHP ) =
IMHO ) -

To ?É Hi -
MY

+1T¥ = 74¥03 -

E. Hint fatty .

= # +
ÉIE ai .
Hi -

consider %/o=q =
¥ ÷qaÉHi +
- MY =D .

-
: ñ=ñÉ*mi
*T¥lo=f 1741104 ¥4 Himiko 'll / of
=
+

= ¥ ¥ -

¥ # ni
,
-

=
¥ (E -
n
) = ¥ -
< o
.

Mi maximize
'ñ? (
and hence 401
. : f = Xi -
Ink .

Ñ
.
: MLE estimator of D= .
 ELOY =

EH ?É* ) -
m) = ti E É×i- Him + MY ]
=
ᵗnÉ Xi ,
-
2M El?É✗:) +
Énñ
=
tie #ix.il -

nai
EIR ) -

END ! varix) = 02 ,
:
Effy ,
= n EH
?
= no} n µ

E (d) Ñ :& is unbiased estimator of 0

.

i. =
.
an
,
?⃝
(a) I = 56.2
,
5=57.9

for I HER
(b) PH=x ) = ✗=o . . .
. -

, .

= PIXFXI ,
k=Xr , . . . .
Xn = Xn ) = PHI -41 ) Parti ) . .

.
PHn=Xn )

X _É* x!✗zi"✗I
"
=

IN LIN =
-

int n I MX
-
In Hi ! . . .tn :)

*Y = -
n -19¥
0%41-1×-1=0 = - n -1m¥ ñ=ñ

%" / ✗ =p -9¥ =
so
,
UH and Inuit attain max

when HI .

,
1--56.1 is the estimator 0ft .
HEIM :
Eti ¥ ,
Hi -

it : nite / Fix iii.Ein

-

+ mi
)
Elif ¢-61T
-

= Var H) = X : E =
✗ +72 .

xi-u.it/vif- nHtXY.E-2X?E.Xi+nx2- E-2X.ni+nx-i- Efnx-i

El?Éx i]
-

=
Éxi
+

EET] -

EI= var(F) EET = ✗ 4- ¥

.

-
:
Etsy =
¥, fax + mi mi x
- -

) =
¥t ,
=
7

i. P is an unbiased estimator of 7 .
(d) ÉLX ] E(✗H"+jY t.nl 7
=
=
=
.

i. ITS also a MLE unbiased estimator of 7 .

Var Gi ) =
var
) =
hint ¥ =

.
=

¥ .

"¥n=%¥&¥ᵗ
"
var = =

Varus ) ≤ var (8) it a ≥ 0 .

of background isotopes ≥
-

mean o .

: I has smaller variance than S2 .

-
'

.
Choose I
 PHI -14 PHI -441 PHI -44

.PL/h- Xn)- f-ox%1 toxi:/. .ftoxnH--O-nHXi. Xn

(a) 1101 =
,
Xv :X } .tn :X n )
. .
= . . .

)¥

.tn/---n1nOt(to-P?4!MXit%FF-iE.4inxi)tr4
1h40) = -

n Ino tf 1) In Hila
- .

%÷%=a =
% ¥ ?Ém×i=o
-
.

i.
8=-1%4 ink
8%910=8 =
¥ + % ¥.mu/H--t
=
¥ In -
In ) =

% < o .

i. MUM and 110 ) attain maximum when 8=9

.
: D= tn ?É-
Inti is a MIE estimator of it .

,
lb Eloy Efta=

Fy ),
Ink ,
:
'

k EIMXH Mtn .
+ lnxn)

Efim] =

ft ☒ to ✗ ¥ ax
consider ff ✗ Pinxdx =

f! lnxd # XP ) "

¥ , Mind ! f ! XP -
"
txdx )
=
¥ tf ! Had
¥ ¥1M / ! p+%
- "
= =

f- ¥ PH .
= + I =
to , -¥p = -02 .

i. Élnif =
t .

-02 =-D .

RV
^

! Xi , Hi "
" ✗ n are independent and
identically distributed .

I Eff ) =
In ④( MH ) -1 Elm xD + ii. + Elin ✗
ni
-

En ) -41M =
0

÷ ITS an unbiased estimator of it .

(a) EE] = E
ftp.yxi/:tnEHi+XrtmtXn )
= Efx) =
p-io.li p ) =p
1- -

t .
Eli ] = 4. np xp .

% ITS an unbiased estimator of pi

lb var 1×7 =
var En ¥ × :) ,

=
¥ [
Var H , ) + Var (X2) + i. it
))
Varin

EE) +0% pl =p
'

:
=
I :p - .

"
Efx) )
"

- : Vara ) =
Efx) -

=p p -

Vakil ¥ # F) PYE
=
.
i.
.
H É×¥Y) =
tn ( Eli ) Eli -

7)
EE] KEN: varli)
- =
K¥ .

E(*×" )
×"
'

: Eli] = :
% np =p .
.

:<
Elif =
P
+ p
'

:
EH" I =
tn ( p -41¥ if +

= b- pltp.lt
.
-

1)
=
kD¥
(d) Etat 4- =
C EH a- x-D =

P¥ .

1m¥ 't"
n
e. i
Pl¥ .

C "
¥1 .
(a) f-(5-5) =
POESY =
Is / {f- Is:X ) dy
=
Est ? ,÷ÉÉ Édx .

=
(E)÷é¥ᵗ .

Els] =
fists Hs =

¥ % ᵗé%s .

East :c
=¥¥¥÷ É %É→E%÷r ,
-

(b) TG :|? #)
"
etat .
/ 9t%ttH =
-

ᵗtᵗ -1 ? t Df!tᵗtᵗÑt}
-
-

=
(2--1)/9 t%ᵗdt .

4--176--41 ? # e- txt
= (7-1×7-2) G- G- D) .
.
.

f? litdt
= ! flit ? = ED !
" T (E) =
CE 1) ! -

,
HE> =
#D !
,
1? (7) =
(2--17177-1)

n -5
-

=
c- -

¥j¥¥☒ =
¥5,4 =
¥ fit,
≈ 1.06 .

n :b : C :

¥ ±É¥- ,
= 3%1=1 % 1.05
"
I.% " I
 Yi follow normal distribution ,

f- Hi
-

y D=
-

arty exp )
E- Eli] =
EIGHT { if =
( ✗it 0 =
9×-1
Var Fli ] = Var -19×-1+4 ;) = 0+02 = or .

-
:
f-Hi Yi
-

-
)=*ie×p(-¥¥ )

ftk-ydflyi-44.FI/n=yn)--KoEe-ErE.H-ex-i
Lll ,
M) =
fly ,
-

-
Yi , Yuya .
.
. .

Yn y a)
-

- =

In 114,04 = % In frog IF -

?É Hi -1*5
,

01T¥ ! ¥ e) ¥44 : -1×-1 'tx-D =

¥ ,É H -1×-1
:

ᵗ"¥ = ¥ +
ÉÉÉH -1*5 :

=
¥ +
¥ Hi -1*5
,

""Y le:p -

- o
, ᵗ%! to :*
= °
.
 É÷¥¥÷
.ir?tn.Eti-. e*5%fa*,-fe.*Y YTki .ei- io-i
E.

,É G -1*1=0
:

%%=¥¥:*
.

I%He :*:-# ¥41M

" "

=i¥n :/ =

:*

¥÷¥¥ ÷!¥¥d= ÷*÷

timidity finding
"" "

*Yf lie:* :¥É✗ico .

i. Million)hndLµÑ) attainmaxhhenf.fi F-

i. ME mentor :
tore .tn?E.H-.-l*ito
 •

(a) f- Hyp ) =
p✗ltPY✗ ,
two or t.pt to if ,

In f- Hip ) = ✗
Mp (H ) MHP)
+

*nt¥p x-p Ei ᵗ"¥¥¥=¥iñ = -

El "¥¥H=¥ÉH+Ép¥HH
P'
"
'
: Eff] = O '

4- pitot 1. P' 4- p ) =
p .

ÉÑ¥¥P] tpup-i-p.plp.tl p¥
'

:
-
: =
.

-P4
lower bound

n÷ _P%ˢ
= = =
.

_pa
(b)
efficiency = = I =
100% .

peggy
(a) f-4) D= * ✗ e¥ Infix;D ) = -21 not MX -

*¥ &IH ¥
"
=
¥ ,
=
-
¥}

1=1*1*-9 ) Efp = _
Fix] =
F- FEED .

EHT-f xf.ie#dx-.fi/fxafoe-Eg- lfHoHie%Ffixe-*dxJ

"
-

=
If ?xatoe¥ )
=
:* #¥1 ! -1%-4×1
'

I.
-0*01? 20 .

= =

t.FM#EY-.2-o-F. w- =o-F- ¥nF1owerbourd- -n.-g- -

i.

.
1101 =
f- HIM ,
Xv :X } . . . ✗ n :X n ) =
f- Hi -44 ) far -44 fun . . . .
=
Xn)

=
#✗ a. xD exp ftp.EY-D
1h40 = -2M not In * ✗i. Xn )"
-

f ?É ,
✗
-1 .

% / o-a.EE Toni] / + a
=
¥ ¥nI=o + .

§ É =

% to :& =
¥ . -
¥ milo :& .
¥ In -1mi co
.

i. MUM and 110 attain Max when Ñ E =

,

A- É is ME estimator of 0

Asymptotic variance of it ≥ =
(b) f-
4) D) To}Ñe¥ fly ;D ) in 31h0 -121h X ¥
-

In
-
-
= = .

*n¥¥! =p +
¥ ,
%% ¥ ¥ =
-

1=1*1*1 ) Éfp_
'
=
=
É -

FEED .

"
-

EH] =

f ? ✗ É✗é¥d✗
=
Epf ? ✗Hfoe¥) .

=
to )¥F-%ñe¥*Y

ftp.f?xEdi-&.zo=3oltromkD.:Ef&MIf#if
=
{

=
¥ _
=p} . 30 =
¥ É ¥
-
-

lower bound =

_n¥ =
% .
1101 =
f- HIM ,
Xv :X 2 ,f- Hi -44 ) far -44 fun
. . . ✗ n :X n ) = . . . .
=
Xn)

↑ Kiki xD exp to ?É x-D

"
= . .

In 401 = -
MMM not In # i. Hi to
?É -1
✗ "
_

,
.

% to a o ¥ ni / - = + *a
=
¥ ¥ ni +
=
o .

8=5
%ⁿ to :& =
¥ ¥ milo :& - .
¥ Gn -1mi E) .
co
.

i. MUM and 110 attain Max when Ñ =

¥ ,

D= is ME estimator .

A4ymptn.io variance of it ≥ =
É÷
 -8€

(c) f-( ; 8)
x = 0×0-1
,
infix ;D) =
Not (0-1) MX

%¥ = to + in × ,
*%¥ =
-6-2 .

É%¥→= Effy =
To .

lower bound =
NIET =
.

"
consider 401 =
farad f-Hi -414 fun -1in ) . .
. = On Hixon XD
.

IMM = nino + d- 1) ¥4M Xi , % Ft ?ÉMM =

*I Ioa It !¥in×i=o: 0 -

É÷*
&¥¥ / a :p
=

% co
,
1101 and MUM attain max when P :& .

Ñ=¥%µ estimator of
: is MIE 0 .

Any mptn.io variance . =

.¥÷µy