STAT 211

Handout 6 (Chapter 6): Point Estimation

= + error of estimation, is an unbiased estimator of if

!= for ever" possible value of . #ther$ise, it is biased and %ias = E

!& .
'ead the e(ample ).2 "our te(tboo*!.
E(ample 1+ ,hen - is a binomial r.v. $ith parameters, n and p, the sample proportion -.n is an unbiased
estimator of p.
To prove this, "ou need to sho$ E-.n!=p $here
^
p
=-.n.
E-.n! = E-!.n, /sing the rules of the e(pected value.
= np . n =p 0f -1%inomialn,p! then E-!=np 2hapter 3!
E(ample 2+ A sample of 14 students $ho had ta*en calculus class "ielded the follo$ing information on
brand of calculator o$ned+ T 5 2 T 5 5 2 T T 2 2 5 S S S T+ Te(as 0nstruments, 5+ 5e$lett
6ac*ard, 2=2asio, S=Sharp!.
a! Estimate the true proportion of all such students $ho o$n a Te(as 0nstruments calculator.
Ans$er=7.2))8
b! Three out of four calculators made b" onl" 5e$lett 6ac*ard utili9e reverse 6olish logic. Estimate
the true proportion of all such students $ho o$n a calculator that does not use reverse 6olish
logic.
Ans$er=7.:7
E(ample 3 E(ercise ).:! + 0n a random sample of :7 components of a certain t"pe, 12 are found to be
defective.
a! A point estimate of the proportion of all such components $hich are not defective.
Ans$er=7.:4
b! 'andoml" select 4 of these components and connect them in series for the s"stem. Estimate the
proportion of all such s"stems that $or* properl".
Ans$er=7.;;38
E(ample ; E(ercise ).12! +
-+ "ield of 1
st
t"pe of fertili9er.
1
!
1
1
2
<
2
1
1

=
n
x x
S
n
i
i
, E-!=
1
=ar-!=
2

>+ "ield of 2
nd
t"pe of fertili9er.
1
!
2
1
2
<
2
2
2

=
n
y y
S
n
i
i
, E>!=
2
=ar>!=
2

1
STAT 211
Sho$
2
! 1 ! 1
2 1
2
2 2
2
1 1
^
2
+
+
=
n n
S n S n
is an unbiased estimator for
2

0t means that "ou need to sho$

+
+
n n
S n S n
E
( ) ( )
2
2
2 1
2 2
1
2 1
1
2 1
2
2 2
2
1 1
2
! 1
2
! 1
2
! 1 ! 1
S E
n n
n
S E
n n
n
n n
S n S n
E
+

+
+

+
+


2 2
2 1
2 2
2 1
1
2
! 1
2
! 1
=
+

+
+

=
n n
n
n n
n
E(ample 4 E(ercise ).13! + -1,-2,?.,-n be a random sample from the pdf f(!=7.41+(!, &1(1,
&11. Sho$ that
< ^
3 X =
is an unbiased estimator for .
0t means that "ou need sho$
=

^
E
.
! 3 3
4
<
^
X E X E E
chapter
=

= $here
3 3 2
1
3 2
1
4 . 7
3 2
4 . 7 ! 1 4 . 7 !
1
1
3 2 1
1

=

+ + =

+ = + =

x x
dx x x X E , 1 1
The standard error: The standard error of an estimator
^

is its standard deviation

.
The estimated standard error: The estimated standard error of an estimator is its estimated standard
deviation
^
^

=
^

s
.
The minimum variance unbiased estimator (MVUE): The best point estimate. Among all estimators
of that are unbiased choose the one that has minimum variance. The resulting
^

is @=/E.
2
STAT 211
E(ample )+ 0f $e go bac* to e(ample 1, the standard error of
^
p
is
n
p p
p Var
p
! 1
^
^

=
$here
! 1 !
! , 1
3
2
var
^
!
p np X Var
p n Binomial X
Chapter iance rulesof
n
X Var
p Var
=
= =

n
p p
n
p np ! 1 ! 1
2

=

E(ample 8+ 0f $e go bac* to e(ample 4, the standard error of
^

is
n n n
X Var
X Var Var
chapter
2 2
4
<
^
3
A
3
A
!
A A
^

= =

=
$here =ar-!=
A
3
A 3
1
!B C !
2 2
2 2

= = X E X E
E-
2
!=
3
1
; 3
1
; 3
1
4 . 7
; 3
4 . 7 ! 1 4 . 7
1
1
; 3 1
1
2
=

+ + =

+ = +

x x
dx x x
E(ample :+ Dor normal distribution,
< ^
x =
is the @=/E for . 6roof is as follo$s.
3
STAT 211
The follo$ing graphs are generated b" creating 477 samples $ith si9e 4 from E7,1! and calculating the
sample mean and the sample median for each sample.
E(ample A E(ercise ).3!+ Fiven normall" distributed data "ield the follo$ing summar" statistics.
Variable n Mean Median TrMean StDev SE Mean
thickness 16 1.3481 1.3950 1.3507 0.3385 0.0846
Variable Mini! Ma"i! #1 #3
thickness 0.8300 1.8300 1.05$5 1.64$5
a! A point estimate of the mean value of coating thic*ness.
b! A point estimate of the median value of coating thic*ness.
c! A point estimate of the value that separates the largest 17G of all values in the coating thic*ness
distribution from the remaining A7G.
Ans$er=1.8:13:
d! Estimate 6-H1.4! The proportion of all thic*ness values less than 1.4!
Ans$er=7.)83)
e! Estimated standard error of the estimator used in a!.
Ans$er=7.7:;)24
;
STAT 211
1.8 1.3 0.8
thickness
Boxplot of thickness
2.4 1.4 0.4
99
95
90
80
70
60
50
40
30
20
10
5
1
Data
P
e
r
c
e
n
t
1.074 AD
!oo"ness of #it
$or%al Pro&a&ilit' Plot for thickness
() *sti%ates + 95, -.
(ean
/tDe0
1.34812
0.327781
() *sti%ates
METHOD O! O"T#$%$%& PO$%T ET$M#TO'
1. The @ethod of @oments @@E!
Iet -1,-2,?.,-n be a random sample from a pmf or pdf. Dor *=1,2,?., the *
th
population moment
of the distribution is E-
*
!. The *
th
sample moment is

=
n
i
k
i
x
n
1
1
.
4
STAT 211
Steps to follo$ + 0f "ou have onl" one un*no$n parameter
i! calculate E-!.
ii! eJuate it to
<
1
1
1
x x
n
n
i
i
=

=
.
iii! Solve for un*no$n parameter such as 1!.
0f "ou have t$o un*no$n parameters, "ou also need to compute the follo$ing to
solve t$o un*no$n parameters $ith t$o eJuations.
iv! calculate E-
2
!.
v! eJuate it to

=
n
i
i
x
n
1
2
1
.
vi! Solve for the second un*no$n parameter such as 2!.
0f "ou have more than t$o un*no$n parameters, repeat the same steps for *=3,?.. until "ou can solve it.

E(ample 17+ Sho$ that @@E of the parameter in 6oisson distribution is
<
x
There is one un*no$n parameter.
The 1
th
population moment of the distribution is E-!= .
The 1
th
sample moment is
<
x
Then
<
x
is the @@E for
E(ample 11+ Dind the @@E for the parameters and in gamma distribution.
There are t$o un*no$n parameters.
The 1
th
population moment of the distribution is E-!= .
The 1
th
sample moment is
<
x
Then =
<
x
but this did not help to solve for an" un*no$n parameter. ,e need to
continue the steps.
The 2
nd
population moment of the distribution is E-
2
!=
2
1+!.
The 2
nd
sample moment is

=
n
i
i
x
n
1
2
1
Then
2
1+!=

=
n
i
i
x
n
1
2
1
Since $e have 2 un*no$n parameters and t$o eJuations, $e can solve for the un*no$n
parameters.
The @@E for and are

n
i
i
x x
x
1
2
<
2
<
!
and
<
1
2
<
x
x x
n
i
i
=

, respectivel"
E(ample 12+ Dind the @@E for the parameters and
2
in normal distribution.
There are t$o un*no$n parameters.
The 1
th
population moment of the distribution is E-!= .
The 1
th
sample moment is
<
x
Then =
<
x
but $e still need to solve for the second un*no$n parameters. ,e need to
continue the steps.
)
STAT 211
The 2
nd
population moment of the distribution is E-
2
!=
2
+
2
.
The 2
nd
sample moment is

=
n
i
i
x
n
1
2
1
Then
2
+
2
=

=
n
i
i
x
n
1
2
1
Then this can be solved for the second un*no$n parameter.
The @@E for and
2
are
<
x
and
n
x x
n
i
i
=

1
2
<
, respectivel"
2. The @ethod of @a(imum Ii*elihood @IE!
Ii*elihood function is the Koint pmf or pdf of - $hich is the function of un*no$n values $hen (Ls
are observed. The ma(imum li*elihood estimates are the values $hich ma(imi9e the li*elihood
function.
Steps to follo$+
i! Metermine the li*elihood function.
ii! Ta*e the natural logarithm of the li*elihood function.
iii! Ta*e a first derivative $ith respect to each un*no$n and eJuate it to 9ero if "ou have m
un*no$n parameters, "ou $ill have m eJuations as a result of derivatives!.
iv! Solve for un*no$n Ls.
v!2hec* if it reall" ma(imi9es "our function b" loo*ing at a second derivative.
E(ample 13+ Sho$ that @IE of the parameter in 6oisson distribution is
<
x
There is one un*no$n parameter.
I=Ii*elihood = p(1,(2,?.,(n! = p(1!p(2!?.p(n! b" independence
=
N
1
1
x
e
x

.
N
2
2
x
e
x

.
N
3
3
x
e
x

??..
N
n
x
x
e
n

=

n
i
i
x
n
x
e
i
1
N

lnI!=
( )

+ ! N ln ! ln
i i
x x n
7
! ln
= + =


i
x
n
d
L d
then
<
x =

7
! ln
2 2
2
< =

i
x
d
L d
then the @IE of is
< ^
x =
The 0nvariance 6rinciple+ Iet
1
^

.,
2
^

,...,
m
^

be the @IELs of the parameters

. Then the
@IE of an" function h
1
,
2
,...,
m

! of these parameters is the function h

.,
2
^

,...,
m
^

! of the
@IELs
E(ample 1;+
8
STAT 211
(() Iet -1,?,-n be a random sample of normall" distributed random variables $ith the mean and the
standard deviation .
The method of moment estimates of and
2
are
<
x
and
n
s n
n
x x
n
i
i 2
1
2
<
! 1
!

=
, respectivel"
The ma(imum li*elihood estimates of and
2
are
<
x
and
n
s n
n
x x
n
i
i 2
1
2
<
! 1
!

=
, respectivel"
()) Iet -1,?,-n be a random sample of e(ponentiall" distributed random variables $ith parameter .
The method of moment estimate and the ma(imum li*elihood estimate of are
<
. 1 x
.
(*) Iet -1,?,-n be a random sample of binomial distributed random variables $ith parameter p.
The method of moment estimate and the ma(imum li*elihood estimate of p are -.n.
(+) Iet -1,?,-n be a random sample of 6oisson distributed random variables $ith parameter .
The method of moment estimate and the ma(imum li*elihood estimate of are
<
x
.
All the estimates above are unbiasedO Some >es but others Eo. $ill be discussed in class!
E(ample 14 E(ercise ).27!+ random sample of n bi*e helmets are selected.
-+ number among the n that are fla$ed =7,1,2,?..,n
p=6fla$ed!
a! @a(imum li*elihood estimate @IE! of p if n=27 and (=3O
b! 0s the estimator in a! unbiasedO
c! @a(imum li*elihood of 1&p!
4
none of the ne(t five helmets e(amined is fla$ed!O
d! 0nstead of selecting 27 helmets to e(amine, e(amine the helmets in succession until 3 fla$ed ones
are found. ,hat $ould be different in - and pO
E(ample 1) E(ercise ).22!+
-+ the proportion of allotted time that a randoml" selected student spends $or*ing on a certain aptitude
The pdf of ( is f(P!=

x ! 1 + , 7(1, Q&1.
A random sample of 17 students "ield the data+ 7.A2, 7.8A, 7.A7, 7.)4, 7.:), 7.;8, 7.83, 7.A8, 7.A;, 7.88.
a! #btain the @@E of and compute the estimate using the data.
2
1
2
! 1 ! 1 !
1
7
2 1
7
+
+
=

+
+ = + =
+

x
dx x x X E
Set E-!=
<
x
and then solve for .
The given data "ield
<
x
= 7.:7 then the method of moment estimator for is
: . 7 1
1 ! : . 7 2
1
1 2
<
<
1

=
x
x

=3
b! #btain the @IE of and compute the estimate using the data.
:
STAT 211
I=Ii*elihood=

= = =
+ = + =
n
i
i
n
n
i
i
n
i
i
x x x f
1 1 1
! 1 ! 1 !


lnI!=

=
+ +
n
i
i
x n
1
! ln ! 1 ln

=
+
+
=
n
i
i
x
n
d
L d
1
! ln
1
! ln

=7 then solve for .
The given data "ield =

=
n
i
i
x
1
! ln &2.;2A4 then the ma(imum li*elihood estimator for is
!! ;2A4 . 2
! ;2A4 . 2 17
! ln
! ln
1
1
^

+
=

=
=
n
i
i
n
i
i
x
x n
=3.11)1

Proposition: /nder ver" general conditions on the Koint distribution of the sample $hen the sample si9e
is large, the @IE of an" parameter is appro(imatel" unbiased and has a variance that is nearl" as small
as can be achieved b" an estimator.
A